<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Floor Drees on Medium]]></title>
        <description><![CDATA[Stories by Floor Drees on Medium]]></description>
        <link>https://medium.com/@floriendrees?source=rss-5fc8bd12050e------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*fuAxdp2aw8Ri_9vT.</url>
            <title>Stories by Floor Drees on Medium</title>
            <link>https://medium.com/@floriendrees?source=rss-5fc8bd12050e------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 05:31:22 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@floriendrees/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com" rel="hub"/>
        <item>
            <title><![CDATA[Are you ready to Git Legit?]]></title>
            <link>https://medium.com/@floriendrees/are-you-ready-to-git-legit-d1010c3d2854?source=rss-5fc8bd12050e------2</link>
            <guid isPermaLink="false">https://medium.com/p/d1010c3d2854</guid>
            <category><![CDATA[meetup]]></category>
            <category><![CDATA[ruby]]></category>
            <category><![CDATA[version-control]]></category>
            <category><![CDATA[http-request]]></category>
            <category><![CDATA[git]]></category>
            <dc:creator><![CDATA[Floor Drees]]></dc:creator>
            <pubDate>Sun, 28 Apr 2019 11:36:25 GMT</pubDate>
            <atom:updated>2019-04-28T11:36:25.999Z</atom:updated>
            <content:encoded><![CDATA[<p>The March Amsterdam Ruby meetup featured talks by Pauline Vos (@vanamerongen), Software Engineer at Werkspot, and <a href="https://github.com/minivan">Ivan Zarea</a>, Lead Architect at Young Capital.</p><p>Ivan’s session was an odd one compared to the type of talks we normally host, and difficult to summarize. Suffice to say it involved (many) Lord of the Rings references, wizard hats and the button on this wonderful web app:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nMgYyCIZ3usRoqITFcy0zQ.png" /><figcaption>thebutton.dev — greatest app of all times</figcaption></figure><p>To explain what happens once we click the button and up the click count, Ivan took us on a magical adventure. You prefer your requests magic-free? I can recommend this gem:</p><p><a href="https://medium.com/devopslinks/http-and-everything-you-need-to-know-about-it-8273bc224491">HTTP and everything you need to know about it</a></p><p>Alright. Let’s unpack Pauline’s talk.</p><h3>Git Legit</h3><p>In case you’re of the audiovisual kind, a version of this talk was recorded at GrowIT conference:</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fsiteproxy.ruqli.workers.dev%3A443%2Fhttps%2Fwww.youtube.com%2Fembed%2FO7LAeqng-SI%3Ffeature%3Doembed&amp;url=https%3A%2F%2Fsiteproxy.ruqli.workers.dev%3A443%2Fhttp%2Fwww.youtube.com%2Fwatch%3Fv%3DO7LAeqng-SI&amp;image=https%3A%2F%2Fsiteproxy.ruqli.workers.dev%3A443%2Fhttps%2Fi.ytimg.com%2Fvi%2FO7LAeqng-SI%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/3d95396dcd1a9a79a1d20d29fc14c8f9/href">https://medium.com/media/3d95396dcd1a9a79a1d20d29fc14c8f9/href</a></iframe><p>Pauline started of with explaining a tomic commits to us: <br>1. Single, irreducible unit — every commit pertains to one fix or feature<br>2. Everything works — don’t break the build on any commit<br>3. Clear and concise — purpose is clear from commit message and description</p><p>And then shared the guidelines she uses for commit messages:</p><ol><li>Separate subject from body with a blank line</li><li>Limit the subject line to 50 characters</li><li>Capitalize the subject line</li><li>Do not end the subject line with a period</li><li>Use the imperative mood in the subject line (“fix” instead of “fixed”)</li><li>Wrap the body at 72 characters</li><li>Use the body to explain what and why vs how (we can judge the “how” from the code)</li></ol><p>… which she adopted from this excellent blog:</p><p><a href="https://chris.beams.io/posts/git-commit">How to Write a Git Commit Message</a></p><p>She then talked about interactive rebase and the commands at your disposal with git rebase -i, including pfor picking commits, e for editing commits, and f for fixup (like “squash”, except that it discards the commit’s log message).</p><p>Another pro tip, Pauline compares git log vs git log --one-line for a quick overview of the changes — which explains the blank line rule from her commit-message-guidlines we covered.</p><p>I loved Pauline’s explanation of binary search. A person drops an egg from a floor of a building and instead of finding out where the breakage occurs — common sense dictates the egg will break as a result of dropping it, always, but let’s humor this one — by increasing the floor with 1 for every experiment, starts in the midde. If it breaks there you know the breakage is between there and the first floor. Our magical eggland person tries the middle of those two. In case the arrow is green there, they try the floor in between that one and the breakage point.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jQvJ4ZI7bFdUwz1N751jew.png" /></figure><p>With git bisect you’ll find the break point in fewer iterations.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8efc5G-NOSFUYuAJZUtC-w.png" /></figure><p>We don’t want <a href="http://commitlogsfromlastnight.com/">commitlogsfromlastnight.com</a> in our application history. “Use commits like you use save in video games. And use descriptive commit messages, so that down the road someone new to your project doesn’t need to turn to the ‘files changed’ tab and code review old information.”</p><p>Rebasing doesnt solve the issues (merge conflicts) you run into when you’re not doing atomic commits.</p><ul><li>git commit --amend changes your commit hash, so you’ll need to force push.</li><li>Pick and choose from individual commits with git cherry-pick [commit hash] .</li><li>… and squash commits with the f prefix to merge it into the previous commit (how often did we push to fix a fix, AMIRITE?)</li></ul><p>For more tips, tricks, and accuracy, I recommend you bookmark the cheatsheet Pauline created:</p><p><a href="http://www.pauline-vos.nl/git-legit-cheatsheet/">Git Legit cheatsheet</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d1010c3d2854" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[My Codemotion Amsterdam 2019]]></title>
            <link>https://medium.com/@floriendrees/my-codemotion-amsterdam-2019-a9d3e2991969?source=rss-5fc8bd12050e------2</link>
            <guid isPermaLink="false">https://medium.com/p/a9d3e2991969</guid>
            <category><![CDATA[developer-relations]]></category>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[open-source-software]]></category>
            <category><![CDATA[diversity-in-tech]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Floor Drees]]></dc:creator>
            <pubDate>Sun, 28 Apr 2019 09:51:37 GMT</pubDate>
            <atom:updated>2019-04-28T09:56:52.876Z</atom:updated>
            <content:encoded><![CDATA[<p>April 2nd and 3rd I spent (at the Microsoft booth) at <a href="https://events.codemotion.com/conferences/amsterdam/2019/">Codemotion Amsterdam.</a> Codemotion is a (very) large tech conference, covering a range of topics, and I believe they welcomedover 1200 visitors.</p><p>Friendly co-workers came by to help me out at the booth, and I could attend some of the sessions. I tried summarizing those below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tTI4LOqV3PIy7rlSNBnksg.jpeg" /><figcaption>Day 0 — the mood</figcaption></figure><h4>Coolblue: From 100 to 1000+ deployments a day</h4><p>Pat Hermens <a href="https://twitter.com/phermens">(@phermens</a>) is the Development Manager at Coolblue. Coolblue is a Dutch company, with a large internal development department. Coolblue is seeing some 1000+ deployments a day, and Pat guided the audience through what enables them to move so quickly, the tools theyuse, and most importantly, the mindset required to enable development teams to deliver at rapid pace.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PSWaSPq0Z8K1uJPrBTlIfQ.jpeg" /><figcaption>Pat Hermens, Development Manager at Coolblue</figcaption></figure><p>Why 1000+ deployments a day even? With 1.13b revenue in 2018 (and 14% growth year over year), and development teams doubling with each year (they’re at 18 teams now) the question of “will it scale” is beyond legit.</p><blockquote>… Apparently we are too much trained to disregard differences in scale, to treat them as “gradual differences that are not essential.” We tell ourselves that what we can do once, we can also do twice and by induction, we fool ourselves into believing that we can do it as many times as needed, but this is just not true.<br><strong>Prof.dr. Edsger W. Dijkstra, from “Notes on structured programming” (1969)</strong></blockquote><p>In a talk titled “Faster to master” from a few years back, was an exploration of DevOps practices being put in place at Coolblue. Pat collected the buillding blocks for a good DevOps practice in a “Faster to master” checklist:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/554/1*XjagRTVWq_baLIVNQrXYKQ.png" /><figcaption>“Faster to master” checklist</figcaption></figure><p>Over time Coolblue ticked off ever more checkboxes, and defined the 4 prerequisites for their dev teams to do their best work:<br>+ Responsibility<br>+ Autonomy<br>+ Ownership<br>+ Failure (addressing it as it is, a learning opportunity)</p><p><strong>Responsibility</strong><br>The Hosting &amp; Deploment (DevOps) team used to be the core blocker for delivery. As the number of software teams grew exponentially, so did the informal networks between teams, exchanging strategies on how to circumvent DevOps.</p><p>By enabling the teams to take responsibility for Ops themselves — and by using linters, and an automated CI/CD process — Hosting &amp; Deployment are now a team just like any other, and no longer bottleneck.</p><p><strong>Autonomy</strong><br>Coolblue has a lot of services, all called Vanessa (something). While services that have been broken apart from the delphi monolith aptly named Vanessa Optima Prime are more fit to use applications like Splunk, Datadog and thus logging, and alerts, Prime took the company to where they are today.</p><p>One team got the honour and the priviledge to make Prime integrate with modern tooling. While they certainly weren’t succesful at the first try, they had the unequivocal freedom to try solutions as they saw fit.</p><p><strong>Ownership</strong><br>At Coolblue they take ownership very seriously. Like a contract, it is well documented who owns what, and the team that owns a topic is free in the tools they want to use. Pat confesses he doesn’t know whether or not Coolblue “does” Kubernetes, or Serverless, because it doesn’t matter.</p><p><strong>Failure</strong><br>Coolblue is serious about error reporting. When a campaign (buy a coffee machine, get a milk frother for free) f-ed up the provisioned space planning for their bike couriers, Coolblue knew about it in the instance it happened.</p><blockquote>Fail early, fail often, but always fail forward.<br><strong>John C. Maxwell, in“Failing forward”</strong></blockquote><p>In a RFO (Reason for Outage), the people responsible are to explain why instances are terminated. In RFO meetings that take place once a month the teams discuss how to mitigate issues moving forward.</p><p>And because we all like a good checklist, Pat shared the one for the “100 to 1.000” journey:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/532/1*75R0XFMazQZPjxkrngDXTw.png" /><figcaption>The “100 to 1.000” checklist</figcaption></figure><h4>Diversity in tech panel: understanding what WE can do every day</h4><p>Diversity in tech: what’s actually being done to make the community more inclusive? We all agree that education, careers, and opportunities in technology should not be limited to any one demographic or sector. At the same time, there are underrepresented groups who are struggling to break into the field.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vOBAzahAl7A-3SmzKN6ziw.jpeg" /></figure><p>MC-ing the panel is Eliza Camber (<a href="http://twitter.com/ElizaCamber">@ElizaCamber</a>), Android Developer at Pixplixity, and GDE (Google Developer Expert). She’s an organizer for the GDG (Google Developer Group) NL and the lead for the WomenTechmakers program in NL. Joining Eliza, Qa’id Jacobs (<a href="http://twitter.com/Qaidj">@Qaidj</a>) is a UX Strategist at Constellation Design. Born and raised in New York and New Jersey (USA), Qa’id moved to Amsterdam in 2013. He has worked with both large, global corporates and small, bootstrapped startups. Using an empathetic design approach, he takes on challenges of class, locale, and context. Lian Li (<a href="http://twitter.com/chimn">@chimn</a>ey42) is a Software Engineer at Container Solutions. She recently co-organized Serverless Days Amsterdam. And Jean Volski is the Global Support Product Lead Developer Tools at Atlassian, with over 15 years experience managing support and services teams across diverse cultures and technical expertise.</p><p>Eliza asked if harmful product decisions could be omitted by creating diverse teams. To which Qa’id said that you never know what a product will do “until the thing is out there” and that making assumptions is hard.<br>Jean: “When you bring people from different walks of life together, the conflict it naturally brings is very healthy. It forces everyone to challenge ideas and assumptions, which pushes the collective forward. But it’s not a silver bullet, if a leader doesn’t know how to manage a multicultural team, that could actually harm productivity.” You can’t just put diverse people in a room and expect it to benefit the bottom line.</p><p>Lian started her career in gaming, which made for a rocky start especially. “At first I tried to be one of the boys, I’d overcompensate in the sexist jokes and devaluing other women.” After therapy, she realized that she hated everything about being a woman in this society. “White men get to represent everyone, but as a woman or person of color you represent whatever box you’re put in, and your actions are put under a magnifying glass.”<br>Qa’id: “We need more and more diverse role models, so that we can mix and match and take away what inspire us.</p><p>Lian on how we need to start (with the) young: “Children are so impressionable, and we can’t expect venture capitalists to do right by our kids. Instead we need to educate the parents so they can support their children’s curiosity.” Qa’id: “All of our actions teach our kids about gender, race, and class. We need to be very conscious about what those lessons are going to be.”</p><p>Qa’id on promoting women: “I like to use the metaphor of someone going to the hospital and their leg is broken. The medical staff will care for their leg and no-one will ask them why they’re being so leg-centric. That’s where the problem is, and that’s what needs to be cured.” I loved how Lian brought up that with the prominence AI is gaining we need to have the conversation about diversity and inclusion now.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*z5AzYKZa2auYXy_so08t_g.jpeg" /><figcaption>Obviously I didn’t make it to Douglas Crockford’s keynote…</figcaption></figure><h4>Breaking into DevRel</h4><p><a href="https://www.twitter.com/thiagoavadore">Thiago de Faria</a> from LINKIT led a panel on Developer Relations with <a href="https://www.twitter.com/degoodmanwilson">Don Goodman-Wilson</a> from GitHub, <a href="https://www.twitter.com/rainleander">K Rain Leander</a> from Red Hat, <a href="https://www.twitter.com/gitbisect">Jason Yee</a> from Datadog and yours truly (from Microsoft).</p><p>Where our roles differ on certain aspects — like what department we report to, if at all — we share many definitions and values. Like how we are an advocate for our product’s and solutions’ users, more so than for our employers. And how integrity and authenticity are paramount. Neither of us would push an unethical product, because no compensation is worth sacrifing our careers over.</p><p>To effectively communicate with your company (or a prospective employer) about what the Developer Relations role entails, we unanimously recommended Mary Thengvall’s ‘<a href="https://www.apress.com/gp/book/9781484237472">The Business Value of Developer Relations: How and Why Technical Communities Are Key To Your Success</a>’.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ds4quqEiK6VPSN1tihj2Ew.jpeg" /><figcaption>From left to right: Thiago, Rain, Don, Jason and me.</figcaption></figure><h4>Chinese food, motor scooters, and open source development</h4><p>Don Goodman-Wilson is Developer Advocate at GitHub, and a philosopher-engineer who has previously worked with/in web security, chatbots, streaming media, embedded hardware, the model railroading industry, and academic philosophy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DxjWilfFTbM_SVmCnevu4A.jpeg" /><figcaption>there are *some* problematic figures within OSS</figcaption></figure><p>Don says OSS is sick, and compromised NPM packages are just a symptom. We strongly identify projects with their maintainers, which is problematic when more often than not these people — overwhelmingly white hetero cis men — are problematic. Case in point: Linus Torvalds. Don suggests we move away from the tight coupling of projects and their maintainers, adopt the Chinese restaurants’ model and share excess demand. Say what?</p><p>Forking is considered a bad thing for power conflicts reasons. But that’s really mainly a culture thing. The opportunity to make copies freely and do with those whatever we please doesn’t play well with our instinct to keep the keys to success to ourselves.</p><p><strong>There’s roughly 4 ways the fork goes:</strong></p><p>1. The death of the fork<br>2. Fork merges with the original (egcs)<br>3. The death of the original (x.org)<br>4. Two competing projects (libc/glibc)</p><p>Chinese food exploded across America in the early 20th century, rapidly adapting to local tastes while also spreading like wildfire. How was it able to spread so fast? A distributed open source-like development model, that’s how.</p><p>Whenever in the US you go to a Chinese restaurant, you’ll get the same exact experience. The north star of Mc Donalds (13.677 locations in the US), is what 41.000 Chinese restaurants achieve with no central coordination. It’s the most successful restaurant model in the world.</p><p>When Nixon went to China in 1965 and ate Chinese food on television, that sparked a huge interest, causing these entrepreneurs to look to alter dishes to appeal to this curious audience, and use locally available ingredients.</p><p>In our culture we have the urge to attribute success to one originator. Instead migration associations in the US that help Chinese immigrants land a job, know where there is demand for a new Chinese restaurants, or chefs. Via these networks restaurant owners share dishes that work with their audience, reducing risk because it’s in their benefit when Americans have a favorable attitude towards foreign food. A powerful open source at work.</p><p><strong>The benefits of decentralization:</strong></p><p>1.Removal of ego<br>2. Democratized access to contribution<br>3. Increased diversity of experience and opinion<br>4. Better localized product-market fit</p><p><strong>Copies are cheap</strong>, says Don. We know this from open source software. <strong>Share excess demand</strong>, and <strong>decentralize production</strong>, to reduce risk and benefit as a community.</p><h4>Ethical questions in software engineering</h4><p>“Software is eating the world. We, developers and architects, are a major force influencing software, technology, and the world it creates. We don’t have the privilege of being unaware of our actions.” Rotem Hermon (<a href="https://www.twitter.com/margolis20">@margolis20</a>) is a Lead Architect for Customer Data Cloud at SAP, and asks us whether we’re ok with relying on privately developed and owned software systems deciding who gets a loan, or who’s killed in favour of sparing another person’s life in case an autonomous car needs to make a split-second decision.</p><p>“Technology is not a get out of jail card.” Facebook , Uber and Airbnb can label themselves as technology companies but that doesn’t exempt them from taking responsibility for the industries they disrupt.</p><p>Rotem encourages us to “keep asking questions” and “not take things for granted”. “There’s a scarcity of IT talent, which means we can influence how our companies move forward.” One thing we should not be doing is simply using more software to fix technology.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mmh_WM0MIuVNmmAn7uvgZA.jpeg" /><figcaption>That’s us at the booth</figcaption></figure><p>I did a talk at CodeDaze in Buffalo last year that covered this topic as well. For a transcipt: <a href="https://medium.com/@floriendrees/start-taking-responsibility-for-the-stuff-you-build-3e06f7d045f4">click</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a9d3e2991969" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Set your GPS to DevOps]]></title>
            <link>https://medium.com/@floriendrees/set-your-gps-to-devops-5a6a3eb15416?source=rss-5fc8bd12050e------2</link>
            <guid isPermaLink="false">https://medium.com/p/5a6a3eb15416</guid>
            <category><![CDATA[pagerduty]]></category>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[continuous-delivery]]></category>
            <category><![CDATA[continuous-integration]]></category>
            <dc:creator><![CDATA[Floor Drees]]></dc:creator>
            <pubDate>Wed, 13 Mar 2019 22:01:40 GMT</pubDate>
            <atom:updated>2019-03-13T22:17:45.897Z</atom:updated>
            <content:encoded><![CDATA[<p>Tuesday March 12 I attended the <a href="https://www.meetup.com/nl-NL/DevOpsAmsterdam/">Amsterdam DevOps meetup</a> at TomTom in Amsterdam, which had a focus on AI and Kubernetes.</p><p>Martijn Siereveld, Director Software Engineering at TomTom, opened saying that TomTom long moved on from putting navigation devices in “your dad’s car”, focusing on maps, autonomous driving, and automotive and enterprise customers. TomTom powers Microsoft cloud offerings (Azure) with location-based services.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5vwgLHYXBX6TmQyG0l69iw.jpeg" /></figure><h4>A DevOps approach to AI</h4><p>Thiago de Faria (<a href="https://twitter.com/thiagoavadore">@thiagoavadore</a>) is the Head of Solutions Engineer for full-service IT provider <a href="https://www.linkit.nl/en">LINKIT</a>.</p><p>A report from MMC, a London-based venture capital firm, recently made the claim that 40% of European firms that are classified as an “AI startup” don’t actually exploit the field of study in any material way for their business:</p><p><a href="https://www.forbes.com/sites/parmyolson/2019/03/04/nearly-half-of-all-ai-startups-are-cashing-in-on-hype/#74804801d022">Nearly Half Of All &#39;AI Startups&#39; Are Cashing In On Hype</a></p><p>Thiago’s mission is to make us all feel more confident in supporting AI and Machine Learning applications. Those that are in fact real.</p><p>You can’t catch bugs in your machine learning algo with console.log(). You can’t “patch” ML models. Instead, Thiago suggests to borrow best practices from DevOps.</p><p>Instead of<br>1. Ops object to Dev using untested frameworks<br>2. Dev adds technical debt to circumvent Ops<br>3. Ops object a bunch more</p><p>… Thiago says Product, Dev, IT Ops, QA, InfoSec, Data Science should all be holding hands and collaborating. That means undergoing a similar culture shift we went through with DevOps:</p><ul><li>Data Scientists want to be “on-call” even less than you and me do. Thiago thinks they should be on the rotation, and this should be a top-down decision.</li><li>Acknowledge we’re dealing with ethics issues (and biased models)</li><li>… and explainability problems. Shouldn’t you be able to explain why someone is denied a loan, even or especially when an algorithm made that decision?</li><li>Creating a safe env where people dare to share their models.</li><li>Creating proxy metrics, because it’s hard to measure efficiency otherwise — you can hardly ask users if your model correctly predicted hotdog / not hotdog, every time.</li></ul><p>“Local Data Science is not a thing, just like works-on-my-machine died with containers.” Beyond CI/CD for ML, Thiago advocates for “continuous evaluation”. Bias, drift &amp; fragility are ML’s bugs, says Thiago. Bias, like <a href="https://www.bbc.com/news/technology-45809919">Amazon’s sexist hiring</a> algo. Drift, as in ML models need to be (re)calibrated, (re)trained. And then there’s the fragility of the input data. And the latter two only surface in Prod.</p><p>Thiago demo-ed <a href="https://mlflow.org/">mlflow</a> (by <a href="https://databricks.com/">DataBricks</a>) and its use for evaluating, packaging, deployment and model serving, and hopes that Uber will open source their machine learning-as-a-service system Michelangelo.</p><h4>Agnostic CI/CD: the end of the line for DevOps?</h4><p>Ruben Visser is a full-time “IT guy” since 1995. He recently started <a href="http://n-ableconsulting.nl">n-ableconsulting.nl</a> and offers Kubernetes / cloud-native consulting.</p><p>CI/CD should cover your application from cradle to grave, says Ruben, be automated, <strong>and</strong> should be environment neutral (hence, “Agnostic CI/CD” ). Ruben showed how with Gitlab &amp; Kubernetes the whole DevOps cycle can be automated for Microservices projects. He walked us through his namespaces (and dev/prod parity), Git and branching, Helm charts, and other personal tweaks.</p><p>Ruben ultimately ends up with a setup that is indeed agnostic-ish. It still utilizes GitLab and Kubernetes, but admittedly doesn’t depend on it.</p><h4>Announcements and discount codes</h4><p>Use “devopsmeetup” (or “dutchazuremeetup” if they limited it) to get 25% off of your <a href="https://serverlessdays.amsterdam/">ServerlessDays</a> Amsterdam March 29 ticket. With “devopsgogoto” you shave 10% off of your<a href="https://gotoams.nl/"> GOTO Amsterdam</a> entree fee.</p><p>And <a href="https://www.devopsdays.org/events/2019-amsterdam/welcome/">DevOps Days Amsterdam</a> is still looking for speakers (<a href="https://sessionize.com/dodams2019/">go to their CfP</a>), and <a href="https://www.devopsdays.org/events/2019-amsterdam/registration/">Early Bird tickets</a> are still — but limitly — available.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5a6a3eb15416" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A poodle is a dog, Ruby is not Rails]]></title>
            <link>https://medium.com/@floriendrees/a-poodle-is-a-dog-ruby-is-not-rails-83f8d1bb4f0e?source=rss-5fc8bd12050e------2</link>
            <guid isPermaLink="false">https://medium.com/p/83f8d1bb4f0e</guid>
            <category><![CDATA[garbage-collection]]></category>
            <category><![CDATA[ruby-meetup]]></category>
            <category><![CDATA[ruby-memory-usage]]></category>
            <category><![CDATA[linux-kernel]]></category>
            <category><![CDATA[mental-models]]></category>
            <dc:creator><![CDATA[Floor Drees]]></dc:creator>
            <pubDate>Sun, 03 Mar 2019 18:27:53 GMT</pubDate>
            <atom:updated>2019-03-03T18:27:53.307Z</atom:updated>
            <content:encoded><![CDATA[<p>The February 25 Amsterdam.rb (<a href="https://www.meetup.com/nl-NL/Amsterdam-rb/">Amsterdam Ruby meetup group</a>), sponsored and hosted by Salonized, ran two talks too good not to cover (but I might be biased).</p><p><a href="https://www.twitter.com/soveran">Michel Martens</a> joined us from Paris, France, to share the key insights that shaped his way of thinking about web development, a lot of which has to do with (mental) modeling. Michel is the maintainer of the <a href="https://github.com/soveran/cuba">Cuba framework for Ruby</a>, and works at openredis (Redis hosting).</p><p><a href="https://www.twitter.com/honglilai">Hongli Lai</a> is the creator of the Phusion Passenger app server. In 2018 Phusion turned 10 years old, and introduced Generic Language Support to its open source and Enterprise offering. Hongli set out to tackle the Ruby memory usage / malloc arena fragmentation problem, and reported on his journey so far.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ybWBk1HLP827qRDmnzc_pw.jpeg" /><figcaption>Michel Martens in front of 60-ish attendees (I’m the one with the yellow backpack)</figcaption></figure><p><a href="https://blog.phusion.nl/2019/01/24/who-takes-out-your-trash/">Last month’s talks</a> by Rain Leander (about Ruby on RHEL — Red Hat Enterprise Linux), and Sanne Kalkman (garbage collection), were the perfect prequel to Hongli’s talk. And Jarra Schirris’ talk at <a href="https://medium.com/@floriendrees/moving-bad-code-into-separate-services-is-just-that-its-not-microservices-a67c3539bd1b">denhaag.rb earlier this month</a> shared similarities with Michel’s musings on mental models. I’m not sure what that all means, and whether these are topics that are indeed very current — or that it’s just a sign that I attend too many meetups and as a result am seeing connections where there aren’t necessarily.</p><p>ANYHOW.</p><h3>Mental Models and Programming</h3><p>“People, when they try to reason, they make systematical mistakes.” Michel doesn’t sugarcoat it. We’re not very capable of understanding concepts that just surpass basic logic.</p><p>if A is true, then B is true<br>A is true</p><p>… we can safely say that B will return true as well. But what happens when B is false, or A is false, or B is true? Chances are you can flesh out all possible scenarios still, but it’s getting trickier. The more complex a system, the more difficult it gets to reason about it.</p><p>A mental model is a person’s <em>representation</em> about how something works in the real world. We humans are economic about what we store in our brain. Like: we’d store what’s true and get forget about what’s false.</p><p>If Michel were to paint a picture of the following requests:</p><p>GET /<br>GET /foo<br>PUT /foo<br>GET /bar<br>GET /bar/baz</p><p>he’d move the directory tree left, requests right, moving the ‘verb’ to the end. Like so:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/1*lAQZ36h3j3fC2SHSohylMA.png" /><figcaption>verbs at the end</figcaption></figure><p>Mental models are private to you, and cannot be shared. Overt models extend our working memory, and can be shared, like the directory tree in the above example.</p><p>Upon discovering Rum (a framework for Rack) in 2009, that allowed to describe routes in a tree. Taking inspiration from Rum, Michel wrote Cuba a year later. Initially he wrote it as a means to share his ideas with colleagues, but then Cuba was adopted as the go-to language for most all of their projects.</p><p>In 2015 Michel started with the development <a href="https://github.com/soveran/syro">Syro</a>. “Cuba allows you to write your code like a tree, but it doesn’t guide the developer to do it “right”.” Indeed the router promotes a much less flexible usage pattern. Modularity is encouraged and sub-applications are supposed to be dispatched.</p><p>Visual representation is paramount, and on a smaller scale, says Michel, even something like indentation tells a story of what lives where.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/674/1*Fz59HLbZF-eBvSmudtT8AQ.png" /></figure><h4>If it’s Rails, then it is Ruby</h4><p>On a whole, Michel recommends using simple models tackling only a part of the logic to make it easier to reason about it. Use overt models, anything that aids a team’s understanding, because that’s the main goal. Educate yourself on the different types of reasoning (deductive; abductive; inductive).</p><p>Audience questions predominantly focused on evolving our reasoning skills, or training our brains to deal better with complexity. Michel: “Some problems have a LOT of steps, but they’ll follow a loop or a pattern and can be easier than a 5 step process that needs 7 models. Like with chess you can improve your mental modeling by learning patterns.”</p><h3>The problem with Ruby’s memory</h3><p>When a proxy server at Phusion making requests to Amazon S3 showed all the signs of a memory leak, Hongli did what every respectable engineer would do. He took to Twitter. As it turns out <a href="https://twitter.com/nateberkopec">Nate Berkopec</a> (author of “The Complete Guide to Rails Performance”) and Heroku both <a href="https://www.speedshop.co/2017/12/04/malloc-doubles-ruby-memory.html">wrote about Ruby’s memory fragmentation issue</a> and offered two possible solutions.</p><p>Either set a magical environment variable (MALLOC_ARENA_MAX=2), or overrode the OS’s memory allocator in favor of a custom one (<a href="https://github.com/jemalloc/jemalloc">jemalloc</a>). While their solutions worked, Hongli isn’t interested in patches, he rather sees stuff fixed upstream.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/231/1*RPVSN-y87a-QgmYacydpuA.png" /><figcaption>Pretty accurate representation of Ruby’s memory bloat</figcaption></figure><p>The problem only manifests itself in multithreaded Ruby apps, and only on Linux. Which is kind of ironic because multithreading is supposed to give us more concurrency with little added memory usage.</p><p>The operating system comes with a library called the memory allocator. This component is what keeps tracks of what memory is in use and what is available. Its API is simple. The malloc function allocates a piece of memory of a requested size. It returns the memory address of the allocation, or an error.<br>The free function frees the memory piece allocated at the given address.</p><p>Ruby uses that API to allocate memory areas for storing Ruby objects. Such an area is called a Ruby heap page. A Ruby heap page is split into equal sized slots. One Ruby object occupies one slot, whether it’s an integer, a string, an array, a class, …of 40 bytes. Most Ruby objects fit in one slot, and when extra data gets stored elsewhere, the allocator will add a pointer. Is that object is destroyed by the garbage collector, then its slot is marked freed, and the external data is freed using the OS memory allocator.</p><p>The Linux kernel’ 4KB blocks are called a page. Hongli calls these “OS pages”, so that we don’t get them confused with heap pages. OS pages allocate ‘heap’, which similarly, Hongli called “os heap” to distinguish them from heap pages. As long as there’s space left, it will fill up from malloc (memory allocation), until it needs to grab new slots from heap pages.</p><p>Sometimes a garbage collection happens. If an entire Ruby heap page becomes unoccupied, then the entire page is released back to the memory allocator. If there’s just few free slots in heap page it as a whole is considered in use. So if and when memory is allocated unfortunately, we’ll need to ask the kernel to allocate an entirely new heap to satisfy more… stuff. Even though there are free slots for Ruby to allocate memory into, as far as the memory allocator and the kernel concerned, they’re still in use.</p><p>The memory allocator has a similar yet very different problem. The memory allocator doesn’t have to release entire OS heaps at once. In theory, it can release any individual OS page. So let’s say we release the 12 KB piece, or 3 OS pages, at the beginning of an OS heap. No problem — in theory. But we run into an issue of an OS heap contains pieces of memory that aren’t a multiple of 4 KB. Let’s say the app frees that 3 KB piece there. There’s still a small part from that 2 KB piece remaining in the same page, and so that individual page cannot be freed. So if we’re unlucky, then we can end up in a situation where there’s lots of free space inside OS heaps, but not many OS pages that are entirely free. Worse. What if there are many holes, but none of them big enough to satisfy a new allocation request? The memory allocator will have to allocate an entirely new OS heap.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/682/1*796m5PPQAiQeOCfdn7mrBQ.png" /></figure><p>A synthetic test — which, Hongli adds, is likely to yield more optimistic results than a production workload — shows App memory usage at 230 MB, where Ruby only knows about 7 MB. That means the 223 MB is occupied by something Ruby has no control over, such as free space in OS heaps, or maybe memory allocated by libraries that Ruby has no control over.</p><p>So why do the solutions Nate and Heroku suggest work? It has got to do with multithreading. When multiple threads try to allocate memory from the same OS heap at the same time, they compete for access. Only one thread can perform an allocation at a time, reducing multithreaded memory allocation performance. The memory allocator has an optimization for this scenario: it tries to create multiple OS heaps and tries to assign different threads to its own OS heap. Most of the time a thread only needs to work with that one, thereby avoiding contention with other threads. In fact, the max number of OS heaps allocated in such a fashion is, by default, equal to 8 times the number of virtual CPUs. So on a 2-core system with 2 hyperthreads each, that’s 32 OS heaps 😱. Both Nate’s blog post and the Heroku article confirm that more OS heaps equal more fragmentation.</p><p>The MALLOC_ARENA_MAX variable reduces the maximum number of OS heaps allocated for the purpose of reducing multithreading contention, and thus it reduces fragmentation. Lacking tools to inspect an OS heap, Hongli wrote his own visualization tool, which involved reading “a lot of non-obvious C code”. What its graphs tell us that there’s lots of free room that malloc knows about, but the kernel doesn’t:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*U8YAyOpRzWQw654JMhZZCQ.png" /></figure><p>Red areas are memory locations in use. Gray areas are free locations, that are not released back to the kernel. White areas are freed to the kernel. To Hongli’s surprise, most OS heaps look like the one on the left: there’s a significant number of of entirely-gray pages without any red inside.</p><p>So even though fragmentation could be an issue, the graph on the left suggests it isn’t as bad as Nate &amp; Co thought. It turns out that the memory allocator only releases OS pages at the end of an OS heap, and even that is only done occasionally. malloc_trim is an API that forces the memory allocator to free all OS pages to the kernel, not just those at the end. Modifying Ruby to call malloc_trim during garbage collection meant reducing app memory usage from 230 MB to 60 MB. MALLOC_ARENA_MAX=2, the solution Nate suggested, trimmed usage to 53 MB.</p><p>It seems like Linux is the culprit, and it’s allocating loads of OS heaps per thread, like 8 times the virtual cpu’s — just to be sure. Modern Linux is written for the most part by Red Hat, with its enterprise customers who care about performance more than about memory. While it’s the right trade off for their customers, it isn’t for the rest of us.</p><p>In conclusion, fragmentation turns out to be mostly a red herring. There’s still gains to be had by reducing fragmentation, but the main problem is the fact that the memory allocator doesn’t like to free memory back to the kernel. To be continued / submitted for PR.</p><p><em>At the next Amsterdam.rb meetup: </em><a href="https://www.twitter.com/vanamerongen"><em>Pauline Vos</em></a><em>, software engineer at Werkspot, will make your day-to-day life as a Git user easier (think: atomic commits and interactive rebase options). And Ivan Zarea, Lead Architect at Young Capital (and creator of Pubquizzes), will talk us through the journey of an HTTP request from your browser to your Rails app and back again. </em><a href="https://www.meetup.com/nl-NL/Amsterdam-rb/events/258731921/"><em>March 19</em></a><em>, at WeTransfer.</em></p><p>*Phusion Passenger (originally mod-rails) was created to simplify and secure Ruby on Rails deployments. Rails is often referred to as ‘magic’ because of all the stuff it takes care of under the hood.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=83f8d1bb4f0e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Moving bad code into separate services is just that, it’s not Microservices]]></title>
            <link>https://medium.com/@floriendrees/moving-bad-code-into-separate-services-is-just-that-its-not-microservices-a67c3539bd1b?source=rss-5fc8bd12050e------2</link>
            <guid isPermaLink="false">https://medium.com/p/a67c3539bd1b</guid>
            <category><![CDATA[ruby-on-rails]]></category>
            <category><![CDATA[monolith]]></category>
            <category><![CDATA[microservices]]></category>
            <category><![CDATA[postpaid]]></category>
            <category><![CDATA[software-craftsmanship]]></category>
            <dc:creator><![CDATA[Floor Drees]]></dc:creator>
            <pubDate>Thu, 21 Feb 2019 21:38:34 GMT</pubDate>
            <atom:updated>2019-02-21T21:38:34.421Z</atom:updated>
            <content:encoded><![CDATA[<p>Last week I attended a<a href="https://www.meetup.com/nl-NL/denhaagrb/"> DenHaag.rb</a> (The Hague Ruby meetup group) event, at <a href="https://www.mobiel.nl/">Mobiel.nl</a> / Websend. Talks by Martijn van Leeuwen, CTO to a 18 people tech team at Mobiel.nl, and Jarra Schirris, who has freelanced with Tele2, Digidentity and Kabisa, were honest and experimental.</p><h4>Mother — the long journey of a 12 year old Rails App</h4><p>Mobiel.nl and Sim.nl combined processed 1.5 million phone subscriptions to date. Bootstrapped and proud, Martijn talks about Mobiel.nl’s 12 year old monolith.</p><p>Martijn acquainted us with the Postpaid vs Phone Fulfilment flow, where Postpaid is just another term for HUGE complexity and a lot of if then statements. Up until a year ago, the monolith still managed most of the Postpaid complexity before it was extracted to another app.</p><p>Before Mobiel.nl there was Studentmobiel.nl - a (and I quote) vastly different time. Focused on Cashback deals and Product-plus offers, Studentmobiel.nl was looking at 100+ other competitors. The app was build in PHP, with simple hosting, and a CSV export to FileMaker Pro Database. From 2004 to 2006 they went from 100 to 400 orders a week, and Martijn realized something had to change because that can’t scale indefinitely. Then he saw the now infamous <a href="https://www.youtube.com/watch?v=Gzj723LkRJY&amp;feature=youtu.be">How to build a blog engine in 15 minutes with Ruby on Rails</a> vid.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qg54G5M7P_RjXX1Us36jTQ.png" /><figcaption>Studentmobiel.nl, December 2004</figcaption></figure><p>Rails meant a dramatic increase in development output, some kind of structure, and a “helpful and fun community”. Martijn converted freelance projects from PHP to Rails in just a few days, which got him excited about trying Rails for a real world app: the new webshop pdawijzer.nl.</p><p>Lacking experience with Rails, Ruby or “real software development”, Martijn used PHP style coding in Rails’ MVC (Model View Controller). Starting in August 2006, the webshop was live within 1 month using Rails 1.0 (Ruby 1.8.5), using FTP upload releases. For a pilot project, pdawijzer.nl was considered a technical success. In another iteration, Martijn added support for Postpaid subscriptions (aforementioned hell), and his new playground went live October 2006.</p><p>May 27, 2007 marked release night for the new and improved Studentmobiel.nl. Though there were “lots and lots updates and bugs”, Martijn had to focus on automating the fulfilment process because of the growth of sales.</p><p>2007–2009 marks the rise of the smartphone (the iPhone 3G and T-Mobile Dream both released in 2008). Studentmobiel saw a steady growth in sales and people, but coincidentally also an increase of features and technical debt. Martijn was still the only developer working on Studentmobiel.nl. He didn’t use Bundler or some other gem management, Ruby and gem installation on the production environment were done manually. The code had no tests, and everything was functional tested by _the one_ developer. Martijn used PHP style coding with “fat views, fat controllers, and fatter models”. As a result, the time to develop began to increase rapidly. Maintenance due because of poor code quality slowed down velocity, yet most commercial goals were achieved.</p><p>Mid 2010 technical debt catches up with Martijn and he realizes he can’t manage it all by himself. He familiarizes himself with software craftmanship practices, and Studentmobiel hires a second developer. Still on Rails 1.2 in 2010, they spend a painful 3 months migrating to Rails 2.3. They introduce RSpec and move to GitHub. Perhaps needless to say few new features are added during this time.</p><p>2011 through 2013 the legacy app still features traces of good intention everywhere, lacking follow-up. Test coverage for cross model observers were virtually non-existent and caused a lot of refactoring havoc. Major commercial changes in the Postpaid market meant they had to ductape new features. The whole organisation was under enormous pressure to match growth, and more often than not the team worked late to keep up with sales and pre-orders.</p><p>The only way to save Mother (Martijn’s pet name for the monolith) was to extract apps and services from it. The team now churned out mobile websites first, to then focus on desktop. And just when they were making progress, the industry decided to add “Toestelbundels” (phonebundles, I guess… Dutch is weird) in 2015. Anticipation legislation that only recently went into effect, T-Mobile was first to comply and everyone else had to haphazardly follow suit. The number of possible subscription combinations grew from 100k to 5m, and the whole setup exploded, as it did for Studentmobiel’s competitors. From the 100+ competitors in 2006, only 20-ish healthy ones were left in 2015. Then 2018 only 3 or 4 online retails could survive. <br><br>Figuring that Mother should not do product management anymore, but instead focus on customers and order fulfilment, Project Postpaid was started to extract logic into new Rails apps. Cleanup work prolonged the process, but is was VERY necessary. “Moving bad code into separate services is just that, it’s not doing Microservices”, says Martijn. The cleanup in mother — some single commits removed over 31k LOC — had a dramatic (perf) effect.</p><p>After the code cleanup Martijn could finally do a Rails 5 migration (from 3.2), 2 years after its release. They’d start with a Rails 5 boilerplate, setup clean and fast specs, move a file, test functionality, and repeat. Tedious maybe, but the release was a as smooth as you can expect a release to be.</p><p>In its current state, Mother’s test suite (50k LOC) runs in ~8 minutes, with a total coverage of ~45% (which is… what it is), and counts 106.401 LOC. Lots of work remains to be done, but at least they’ve set themselves up for success.</p><p>In hindsight, Martijn wishes he had invested in code quality once the MVP proved right. Looking back, he tells us to not let commercial growth be the only metric of success for a company. He also evangelizes investing in your personal development skills (software craftsmanship) and that of your team.</p><h4>Write software you understand, stupid</h4><p>Jarra Schirris’ talk —so chockful of book references he should’ve gotten himself an affiliate deal with Amazon — touched on a topic I’m super interested in: reducing software complexity.</p><p>With today’s writing software using software (and creating ever more layers of abstraction), we’re out of touch with the stuff that drives our products. Toyota’s software did exactly what it was told to do. The reason it failed (and hurt human beings in the process) is that it was told to do the wrong thing.</p><p>Although he didn’t quote directly from the article, Jarra’s talk reminded me of “<a href="https://medium.com/the-atlantic/the-coming-software-apocalypse-4ffb43f3b288">The Coming Software Apocalypse</a>” by James Somers (for The Atlantic): “Because it can be changed cheaply, software is constantly changed; and because it’s unmoored from anything physical — a program that is a thousand times more complex than another takes up the same actual space — it tends to grow without bound.” And: “Software failures are failures of understanding, and of imagination.” Such a powerful piece!</p><p><a href="https://medium.com/the-atlantic/the-coming-software-apocalypse-4ffb43f3b288">The Coming Software Apocalypse</a></p><p>ANYWAY. Jarra did reference Paul Graham (and the prophetic <a href="http://www.paulgraham.com/hp.html">Hackers and Painters</a>), and Alan Kay, o.a. The latter often philosophizes about the supposed simplicity and scaling of software solutions. Like when he talked about taking a complex idea and boiling it down to its most basic components, so that it would fit on a t-shirt. The first Lisp interpreter should be seen as an example in this class.</p><p>As technology grows more complex, its behaviour mimics the vagaries of the natural world more than it conforms to a mathematical model. Just read “Overcomplicated: Technology at the Limits of Comprehension”, by Samuel Arbesman, Jarra says.</p><p>“The Art of Logic — how to make sense in a world that doesn’t” by Eugenia Cheng teaches us that there are two ways to make something simpler. We can forget parts of it, or we can become cleverer so that things that used to seem incomprehensible become clear to us. The problem with that, says Jarra, is that George Miller aleady claimed in 1958 that the percentage of people who can hold anything beyond nine chunks of information in short-term memory is essentially zero. It’s actually more like 7 chunks.</p><p>While I agreed with a lot of what Jarra was telling us, if all stuff mentioned is required for writing good quality, understandable code, then why at the same time have we’ve been taking entry barriers away to allow the pool of people able to churn out software to grow? Didn’t <a href="https://www.youtube.com/watch?v=zKyv-IGvgGE">DHH tell the audience at Railsconf 2018</a> that people not needing to know how MySQL even works in order to enter the field is a good thing?</p><p>“While many developers kick-started their careers with the 15-minute-blog tutorial Martijn mentioned, I don’t believe it’s necesarily apt for longrunning projects with all its magic, autoloading and enthusiastic inheritance.” To illustrate his point, Jarra later suggested I’d watch “<a href="https://www.youtube.com/watch?v=YX3iRjKj7C0">What Killed Smalltalk Could Kill Ruby, Too</a>”. I see how people who don’t familiarize themselves with topics like security, decoupling, or referential transparency, yet mess around with critical infrastructure, at best set the industry up for failure, or worse yet might output harmful products.</p><p>Jarra pledges to do his part in making sure we don’t lose control of the technological advances of our time, and to translate changes to a curious audience of novice coders.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a67c3539bd1b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[KaaS  & K8s Powered Go Modules]]></title>
            <link>https://medium.com/@floriendrees/kaas-k8s-powered-go-modules-cc264ec56669?source=rss-5fc8bd12050e------2</link>
            <guid isPermaLink="false">https://medium.com/p/cc264ec56669</guid>
            <category><![CDATA[amsterdam]]></category>
            <category><![CDATA[cloud-native]]></category>
            <category><![CDATA[kubernetes]]></category>
            <category><![CDATA[meetup]]></category>
            <category><![CDATA[go]]></category>
            <dc:creator><![CDATA[Floor Drees]]></dc:creator>
            <pubDate>Wed, 20 Feb 2019 15:40:02 GMT</pubDate>
            <atom:updated>2019-02-20T17:40:47.716Z</atom:updated>
            <content:encoded><![CDATA[<h3>KaaS 🧀 &amp; k8s Powered Go Modules</h3><p>Yesterday I attended the <a href="https://www.meetup.com/nl-NL/Amsterdam-Kubernetes-Meetup/">Amsterdam Kubernetes/Cloud-Native Meetup</a> — part of the Cloud Native Computing Foundation (CNCF) groups on Meetup.com — with talks by Alexander Charykov, Innovation engineer at True, and Developer Advocate at JFrog, Leon Stigter.</p><p>There I met an old friend, who’s now working at True, and got a tour of the (gorgeous) offices before the meetup started.</p><p>Before the first talk <a href="https://twitter.com/bongo">Alessandro Vozza</a>, Senior Software Engineer at Microsoft and CloudNative Foundation ambassador, announced that last year’s <a href="https://k8sday.github.io/">Kubernetes Day</a> will have a sequel in June. <a href="https://cloudnative.amsterdam/">Cloud Native Community Day Amsterdam</a> will take place June 14, and Alessandro recommended signing up for the <a href="https://kubernetes.slack.com/">Kubernetes Slack</a> in order to stay informed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*aFPYIBeUgnYLcmmUQ94gcg.jpeg" /><figcaption>Alessandro Vozza at the Amsterdam Kubernetes/Cloud-Native Meetup</figcaption></figure><p>Another event to look forward to is the Hamburg <a href="https://www.containerdays.io">Container Days</a>, taking place June 24–26.</p><p>Alright, enough with the infomercials already, on to the talks!</p><h4>Building Kubernetes as a Service</h4><p>Alexander works on what’s to become True’s Kubernetes platform. Among other requirements for the Kubernetes as a Service (KaaS), Alexander needs it to be able to deploy applications from the get-go, it needs to allow for Kubernetes upgrades, it needs to have high availability (and serve multiple masters), and it needs single sign-on (SSO) for his colleagues’ accounts.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZQs_ltxEEnpJ4Xmo4jivuw.jpeg" /><figcaption>Alexander Charykov, Innovation engineer at True</figcaption></figure><p>Reasonable expectations for a managed containers service provider, according to Alexander, include:<br>- Pipeline for deploying applications with CI/CD<br>- Pre-build popular CMS / Framework or customer-specific templates<br>- Security audit of containers and applications (<a href="https://medium.com/@floriendrees/dev-sec-ops-meetup-on-kubernetes-containers-a-summary-9684955d7ffb">because this</a>)<br>- Unified monitoring and logging<br>- Tracing of requests<br>- Automatic scaling</p><p>Considering Kubernetes deployment options, you can choose to be your own provisioner, which means a lot of implementation work. While Kubespray is an official Kubernetes project, and Ansible based (== easy to extend), it offers no interactive UI, no multi-cluster support and no Kubernetes upgrades. Rancher on the other hand, does offer a simple UI, multi-cluster support, Kubernetes upgrades and SSO support. And might you still get stuck, Rancher offers Enterprise support. Not to say the platform is without its limits. Its not ideal for tuning Kubernetes services, and it’s hard to maintain more than several clusters.</p><p>The current config on the left, and the envisioned setup on the right with a True managed cluster with functionality wrapped in some namespace, and another k8s cluster coexisting next to it:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/886/1*iuWc-ZGLfFpeDKVuxEfkag.png" /></figure><p>We’ll have to wait a little longer, but I’m sure we can look forward to a nice little (big) launch party at True’s offices.</p><h4>Building a Kubernetes Powered Central Go Modules Repository</h4><p>Leon is a Developer Advocate at JFrog. As an outspoken supporter of developers Leon advocates for the needs of developers within JFrog and helps to drive open source tooling like their KubeXray and GoCenter. Sharing his hopes and dreams for the Kubernetes powered central Go Modules repo he starts with a quick history of Go, which is neat for the attendees less familiar with the language and its… quirks.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2IoFLQfcezkvXl0xpt820Q.jpeg" /><figcaption>Leon Stigter, Developer Advocate at JFrog</figcaption></figure><p>Milestones include Go 1.0 as a long term stable release in 2012, v1.5 in 2015 where all of Go is written in Go and support for vendoring is added, and v1.8 in 2017, the introduction of Go plugins. One thing that’s blatantly missing from the release notes and subsequently from Golang up to that point, has been dependency management.</p><p>Then in 2018, with v1.11, Go Modules are added. Modules and vendoring are considered Golang’s official way of doing dependency management. In forking code you trust a person on the internet whose library you’ve been using, to keep their package up. Instead with vendoring you make your own copy (or a mirror) of the 3rd party packages you depend upon, protecting your app from maintainers deleting critical infrastructure willy-nilly.</p><p><a href="https://www.gocenter.io">GoCenter</a> is such a central Go Modules repository, and it’s build on top of Kubernetes being the defacto orchestration platform. At JFrog, Kubernetes runs on the Google Cloud Platform, and deployments are done using <a href="https://hub.helm.sh">Helm</a>. JFrog uses Jenkins for CI, and uses promotions to make sure that requirements are met for every stage of the pipeline. <br><br>At any given time they’ll have 5 clusters running, one for production, 1 for development and several for testing. Following Twelve-Factor Apps, these 5 clusters are very similar (dev/prod parity). Spinning off of GoCenter are several microservices, all but one written in Go (unsurprisingly, its UI is written in Node.js). GoCenter survived load testing with 10k users with 100 requests per second hitting their UI, and the backend loading 10k modules. Using FluentD, Prometheus and Grafana, container health is closely monitored, and team uses Kube-Slack to keep in touch with their nodes.</p><p>Next JFrog wants to add ChatOps to GoCenter, using Slack to allow teams more control. And they’re hoping to hear from the community in order to improve the product.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cc264ec56669" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The data-driven approach to writing developer content is wrong]]></title>
            <link>https://medium.com/@floriendrees/the-data-driven-approach-to-writing-developer-content-is-wrong-5026fcb7f9f9?source=rss-5fc8bd12050e------2</link>
            <guid isPermaLink="false">https://medium.com/p/5026fcb7f9f9</guid>
            <category><![CDATA[developer-relations]]></category>
            <category><![CDATA[tech-writing]]></category>
            <category><![CDATA[devrel]]></category>
            <category><![CDATA[content-developers]]></category>
            <category><![CDATA[data-driven-marketing]]></category>
            <dc:creator><![CDATA[Floor Drees]]></dc:creator>
            <pubDate>Fri, 08 Feb 2019 11:01:21 GMT</pubDate>
            <atom:updated>2019-02-08T11:01:21.862Z</atom:updated>
            <content:encoded><![CDATA[<p>Did you watch Nexmo’s Martyn Davies DevRelCon London talk “<a href="https://devrel.net/developer-marketing/design-your-content-strategy-with-data?mc_cid=3a8cdf3e05&amp;mc_eid=a5f9b97772">Design your content strategy with data</a>”? Well I did. And while there’s a lot of lessons to be learned from his talk, identifying what post will “take off” is (sadly) still largely gut feels.</p><p>I don’t want to spend my time writing a ton of content, and then digging into analytics with scientifically irrelevant sample sizes, to see what performed best and then write a bunch of more posts like it — regardless how strongly I feel about the topic. I would much rather delight a few readers, and leave the generic posts that semi-engages a mass audience to other people. This isn’t me being bitter (really). I have written page 1, Hacker News, “viral” stuff. But clickbait-y articles attract an audience susceptible of clickbait-y content, and that’s not the reader-following I’m after.</p><p>The best indicator for future following is past following — if and when you limit yourself to data-driven content creation. Consistency in topics is something you’d expect from online marketing media, like newsletters. While a fantastic tool for retention, developer relations folks develop in terms of interests and talking points, much like the people they’re connecting with. I dare you to point me to one person in tech that isn’t continuously updating her knowledge. The data-driven approach banks on the belief that we can churn out content that is useful and interesting for our target audience indefinitely, without experimenting with new themes. I am convinced that notion is false. What’s more, that approach won’t grow your audience, nor add diversity to your database.</p><p>That’s why I’ll look at analytics, but only because that’s what I’m expected to do / because we pay for some service (👋 boss). Instead have some three go-to heuristics that I have found to <em>guarantee technical content</em> that is interesting, and helpful to my target audience. These include buddying up with support, summarizing events that I attend, and sitting in on daily scrum events to find out about upcoming features and releases, among other things.</p><h4><strong>Support is a gold mine</strong></h4><p>I periodically check Stack Overflow to analyse support requests. Chances are that if someone took the time to report an issue, there are others (“others” totally being a scientifically valid quantity) experiencing the same thing. Then I’ll write articles with the aim to ‘unstuck’ and delight community members researching the issue. For me, a central part of developer relations is empowering people, which means removing roadblocks and making the product work better for our developers.</p><p>To the same end, one could also type the company’s name in a search engine and analyse the search queries.</p><h4>Summarize community events</h4><p>I’ll also do a lot of recaps of community events, like conferences, webinars and meetups / user groups. Especially the latter is relatively low effort. I’ll walk you through how I approach this kind of content creation, from preparing for a meetup, to note taking, and ultimately promotion.</p><p>As to the <em>why</em>: I feel like summarizing conferences and meetups is a service to the community. Effectively you’re unlocking content, and making it more accessible. As a side effect, my own learning process accelerates by summarizing, and I like to think I get better (and quicker) at distilling a talk’s message with every event. I’m creating a knowledge base of topics and speakers, as well as audience questions that I can use for my “own” events.</p><p>Also: nobody else is doing it, so you’d be hard pressed to find competing articles. Monopoly y’all.</p><p><strong>Preparation is everything</strong><br>Ahead of a conference I’ll have a look at the schedule. If it’s single track I’ll just copy the schedule into a markdown file. But I guess you could use HTML, or a plain text file as well, depending on your CMS. I just like the easy syntax highlighting markdown gives me.</p><p>If there’s more than one track, I’ll carefully select the talks I will want to see, and create my itinerary (including info on the starting times and rooms). Next I’ll copy in speakers’ bio’s, along with Twitter and GitHub handle. I might even look these people up on LinkedIn to see what their background is like. Sometimes I’ll create a (public) List on Twitter, with all the speakers in it, as well as the event’s account.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tGg2593IiCgglXpTJ4FCoQ.png" /><figcaption>How I’d prepare for Frontend Developer Love conference in Amsterdam next week</figcaption></figure><p>Hot tip: conference and meetup schedules are subject to change. Hence I’ll only start this process a few days prior to the event. And I always check Day 0 if my information is still correct.</p><p><strong>Note taking</strong><br>I’ve read all the talk descriptions, so I know when the talk “starts”. Until that point I’ll know if this is a talk where I’ll need to take notes, or where I can take pictures of slides, judging by the presentation style. Some speakers will tell you in advance when they plan to share their slides. If they don’t, and the talk was too high level for me, I’ll often just ask for a link to the slides, in person, or on Twitter.</p><p>After the conference I’ll upload the pictures to some cloud storage, and fill in the gaps in my notes.</p><p><strong>Promotion</strong><br>In case I’m covering a meetup I’ll ask the speakers if I correctly summarized their sessions before sharing the post. I genuinly want to get it right. But usually that also means they’ll share my post once I do. In a tweet about the post I’ll @-mention the speakers, the meetup’s account, and relevant hashtags.</p><p>A conference will have more speakers, meaning more opportunity for likes and retweets. But I’ll take the intimacy, improv and authenticity of meetups over fancy big conferences any day.</p><p>I’ll also privately share the link with people that are subject matter experts for the topics covered, or that know the people speaking. They’re bound to share as well.</p><h4>Attending stand-ups for content discovery</h4><p>I work in small teams where I can easily sit in on the daily stand-up meetings. There I’d hear about:</p><ul><li>Upcoming features and releases</li><li>Challenges engineers faced and how they solved those</li><li>Engineers’ take on trends</li><li>Find projects to cooperate on / that could use gathering of community input</li><li>Contributions to open source projects, whether our own or by other community members</li><li>Working with other/new technology which might have DevRel people advocating for it, whom I can partner up with for content</li></ul><p>I’ll then send the engineer owning the topic a list of questions that at the same time is the outline of the post already. For example:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CVl_FgVKr7UcPwiG2tM2zQ.png" /><figcaption>Don’t save all this stuff on your desktop, silly.</figcaption></figure><p>With her answers — and a lot of search queries and Docs-reading — I’ll write a draft, which I’ll send back to check if what I wrote makes sense. Whenever I work on something with an engineer, her name will ultimately end up on the post.</p><p>From what I’ve seen, over time people expect you to want to write something about the stuff that surfaces in these meetings. And to my pleasant surprise will sometimes have prepared something already. When that happens I could just burst of joy. All I want to do is empower these people that are all so much smarter than me, as in the end they are the people behind the product.</p><p>This isn’t a best practices post, but rather my personal approach to writing developer content. I’d love to hear what you think, and what your go-to strategy entails, in the comments below👇</p><p><em>Huge thanks to </em><a href="https://medium.com/u/bf3e8bb0636b"><em>Don Goodman-Wilson</em></a><em> for proofreading this article.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5026fcb7f9f9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Git Merge Brussels 2019, a summary]]></title>
            <link>https://medium.com/@floriendrees/git-merge-brussels-2019-a-summary-50e942f06497?source=rss-5fc8bd12050e------2</link>
            <guid isPermaLink="false">https://medium.com/p/50e942f06497</guid>
            <category><![CDATA[github]]></category>
            <category><![CDATA[monorepo]]></category>
            <category><![CDATA[conference]]></category>
            <category><![CDATA[ash]]></category>
            <category><![CDATA[git]]></category>
            <dc:creator><![CDATA[Floor Drees]]></dc:creator>
            <pubDate>Wed, 06 Feb 2019 14:23:09 GMT</pubDate>
            <atom:updated>2019-02-06T15:10:18.340Z</atom:updated>
            <content:encoded><![CDATA[<p>Courtesy of GitHub (or: your friendly DevRel Don Goodman-Wilson), 3/4 of the Amsterdam Ruby meetup organizer team was able to attend the annual Git Merge conference, which took place at The Egg in Brussels February 1st. Carrying a serious amount of snacks (and coffee), we <a href="https://twitter.com/FloorDrees/status/1091245686383882240">headed for Belgium</a> well before sunrise, as the weather predictions weren’t necessarily in our favor.</p><p>Below is me trying to summarize my highlights of the event.</p><h4>Scaling repositories</h4><p>More than one talk touched on the topic of mono-repos (consolidating multiple directories). Johan Abildskov (<a href="http://twitter.com/randomsort">@randomsort</a>) is a trainer &amp; consultant at Praqma, helping companies in Scandinavia move towards Git, CD, and a DevOps mindset. Johan introduced available tools and considerations for deciding on the repository structure (Mono vs Many) that’s just right for you.</p><h4>Bridging the gap: transitioning Git to SHA-256</h4><p>Brian M. Carlson <a href="http://twitter.com/bk2204">(@bk2204</a>) is a Git Ecosystem Engineer at GitHub. Git has long used SHA-1 to identify objects in its datastore, but SHA-1 is considered vulnerable to collisions, where a malicious actor could create two different objects with the same hash. Such an attack still requires lots of resources, but Git decided to make the jump to SHA-256.</p><p>SHA-256 is presently considered secure, its hash fits on an 80-column screen, it offers 128-bit security, and it’s supported in all major crypto libraries with good performance, as well as in hardware on AMD and ARM (soon Intel as well).</p><p>Regarding the transition plan, Git wants to minimize the disruption for its users, and preserve interoperability whenever possible. Users get to decide when to transition (independent of others), and the new signed objects will make use of a stronger hash algorithm.*</p><p>I kinda want to share the Implementation Design Guidelines here:<br>- Everything in the .git directory uses the same algorithm<br>- Everything producing Git data uses the .git directory algorithm<br>- Fetch and push can negotiate an algo to use<br>- All other outpost uses the same algo (even output to other Git processes)<br>- Conversions between algos use git fast-export and git fast-import<br>- Binary data formats use a fixed-length 4-byte algo identifier<br>- Text data formats and command line use SHA-256 or SHA-1<br>- Works for future algos as well</p><p>SHA-256 support is not upstream but available for tinkering in a branch. For more information, as well as the staged transition plan: <a href="https://github.com/git/git/blob/master/Documentation/technical/hash-function-transition.txt">click</a>.</p><h4>Git protocols: still tinkering after all these years?</h4><p>Brandon Williams started his career at Google where he made many contributions to Git. He’s now a Software Engineer at Facebook using Rust. And he wants to examine recent changes to the Git protocol focusing on the introduction of protocol v2 with us.</p><p>The issue with the original protocol is that the server’s ref-advertisement includes all refs. The Chromium repository has over 1 million branches and tags, which means that potentially 10s of MBs are sent per request, … and subsequently ignored.</p><p>However, transitioning is no walk in the park. Backward compatibility (using existing URLs) is complicated as older servers will ignore additional information (the stuff that comes after \0) send.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/620/1*NerjAdgqKOwHZGLdJpnoTw.png" /><figcaption>Sorry, I just had to.</figcaption></figure><p>Cloning the Chromium source using the original protocol (v0) took 10.08s. Using protocol v2, it took 2.33s — roughly 4x faster! With well over a million references, 4x speed is non-trivial.</p><p>Other improvements include:<br>- The mechanism to switch protocols<br>- Server-side filtering of references<br>- More extensible and will allow future improvements<br> — CDN offloading + resumable clones<br> — Rebase on push<br> — Remore grep/log for patial clones</p><h4>Native Git support for large objects</h4><p>Terry Parker is a manager of the git-core and Git server teams at Google and contributes to the JGit open source project. Google hired him 11 years ago to work on Eclipse. Terry knows all too well that large binary objects pose a special challenge for Git. But Git’s new partial clone feature and a new proposal to use content distribution networks can help.</p><p>So how large are we talking here? .bin type files are definitely large. And VM images. And APK files. Certainly files over 1MiB are large. But really, all that’s considered “painful” is too large. Native support comes in the form of “partial clones” (i.e. lazy downloading) and it’s adaptable so that clients and servers both can make the tradeoffs they need, while Git protocols understand and participate in the management. Where large clones previously tied up server resources, limiting ability to serve new customers, partial clones take care of the initial filtering, and (lazy) download on-demand, using CDNs (Content Distribution Network).</p><p>Mono-repo with plenty binary files? The Git Project has you covered (“in a release near you”).</p><h4>Git for games: current problems and solutions</h4><p>John Austin <a href="http://twitter.com/kleptine">(@kleptine</a>) has been making games for nearly 12 years and has previously worked at Google, and Microsoft. He founded and currently leads A Stranger Gravity studio.</p><p>Git is the source control system of the modern era. Yet, the vast majority of AAA game studios still use Perforce, SVN, and other more traditional systems. It’s not for a lack of desire, rather, game developers have a unique set of constraints and workflows that make Git unsuitable for the task.</p><p>The game development workflow is infamous for its terabytes of files at the head revision, and loads and loads of binary files. In fact, most changes made are on the assets side (predominantly .bin files), while Git kinda assumes all your files are text.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/199/1*OU9tg0X2fAZPE_cyowPAGg.png" /></figure><p>Repository bloat is a pain, and binary file conflicts are lost work. Git can’t properly diff or compress binary files, and it can’t merge binary files either.</p><p>A spreadsheet detailing who ultimately owns what files and when changes were made to it (what’s the latest version) doesn’t scale to more than a handful of developers or files.</p><p>Also, isn’t this why we built source control?</p><p>Even with the most advanced neural network in the world you can’t merge art. So if you can’t merge, the best you can do is prevent binary conflicts before they occur. An early warning system of some sorts.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/215/1*EIl1Sgt_jESn5l1fGaid2Q.png" /><figcaption>Error: Commit does not descend from existing work done on file.bin</figcaption></figure><p>But such a warning system can only add contraints on the Git level (i.ee. once you commit changes to a local repo), even if John would love to caution before, locally. The <a href="https://github.com/Kleptine/gitglobalgraph">Git Global Graph</a> project is designed to solve the above challenges without compromising on the fundamental structures and benefits of Git, with file “locking” encoded in the branching model. A pre-commit hook checks the Global Graph for conflicts. Managing un-mergeable files, in Git Global Graph binary files are paths, and a commit is valid if it descends from every commit touching the file.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/719/1*E9tUJKipiqGrgW4fAwvylQ.png" /></figure><p>Git Global Graph is open source, and written in Rust. The project still lacks work on divergent branches and file system layers, but contains a HTTP API for queries, as well as standard Git authentication.</p><h4>The art of patience: why you should bother teaching Git to designers</h4><p>Belén Barros Pena (<a href="http://twitter.com/belenpena">@belenpena</a>) is an Interaction Designer for software engineering tools, and free / open source software. While working as the only designer of the Yocto Project, her Linux-engineer coworkers “mustered the patience” to teach her Git. As a designer, learning Git made her more independent and more useful to her development team.</p><p>Explaining what she does, Belén says its her responsibility to build <em>software that makes sense to the people who use it</em>.</p><blockquote>“A mental model is what the user believes about the system at hand (…) A mental model is based on belief, not facts: that is, it’s a model of what users know (or think they know) about a system.”</blockquote><blockquote><strong>Jakob Nielsen, Mental Models</strong></blockquote><p>Belén: “A GUI (like GitHub) should reflect the way the people think, not how the system works. People that have never used Git, don’t have a mental model of Git.”</p><p>Belén was taught Git on a <em>need-to-know basis</em> by her coworkers who <em>avoided Git jargon</em> (or: “incomprehensible language like branches”), nor bothered too much with the concepts like the difference between git push and git fetch. She encourages aspiring mentors to <em>do things with</em> — never for — their mentees, so they create muscle memory. Make sure they take notes, keep a <a href="https://opensource.com/downloads/cheat-sheet-git?utm_medium=Email&amp;utm_campaign=weekly&amp;sc_cid=701f2000000RRDfAAO">cheat sheet</a>, and use the <em>command line instead of a GUI</em> (because of the aforementioned mental model disconnect).</p><p>Belén: “The wall between developers and designers is not made of bricks, but of attitudes. Designers should learn Git as a design material. And they need to know just enough to make sound design decisions. If we want more designers participating in FOSS, we need to help them to do so.”</p><h4>Version control for law: Posey Rule in the U.S. Congress</h4><p>Ari Hershowitz <a href="http://twitter.com/arihersh">(@arihersh</a>) is the Director of Open Government for Xcential Corporation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/508/1*kOtWvbvsD27bTIs6Q1M6pg.png" /><figcaption>Stealing Ari’s pun here.</figcaption></figure><p>A new rule in the U.S. House of Representatives (the ‘Posey Rule’), requires redlined prints comparing documents before and after changes in Committee.</p><p>Unfortunately, the solution isn’t as straight-forward as just using Git:<br>- Changes are made by amendments, not versions.<br>- Changes are acts (over centuries, thousands of pages), not repositories.<br>- Standard diff doesn’t work, because:<br> — Matching ‘the same’ section requires semantic judgement.<br> — We’re talking moves and changes in 1000-page bills. <br> — Meaningful grouping of changes is challenging.</p><p>As a lawyer who codes, Ari works for the U.S. House of Representatives to build document comparison software that works for law, can track changes in law and ultimately will be able to show what the law was at any point in time.</p><p>His team designed an open standard XML for legislation: the United States Legislative Markup (USLM), converted the U.S. Code to USLM, and continues to build tools for machine-readable amendments and references, machine-executable instructions and legally-relevant diffs.</p><h4>Git, the annotated notepad</h4><p>Aniket Subhash Kadam <a href="http://twitter.com/aniketsmk">(@aniketsmk</a>) is an independent consultant who’s been around the startup block “a few times”.</p><p>Aniket: “Ever forgotten what you were working on the night before and needed a few minutes to get back context?” YES. Aniket uses Git-Cola to keep track of a handful of files only, and check those diffs to get back to where he left off. But that is only as useful as your Git discipline.</p><p>For quit context reloading, more working memory and enhanced focus, always be committing:<br>- Commit in small ‘logical units’ (think: refactoring, a set of functions used together in a feature, one complex function) / any set of changes that make sense to view together.<br>- Read every diff line-by-line before commiting, and check if it passes personal code review.<br>- Use descriptive titles to help with code review.<br>- Take the time to refactor the best you can before committing.</p><h4>Git &amp; version control in the enterprise: a panel conversation with Atlassian, GitHub and GitLab</h4><p>Having the 3 version control system vendors on stage was definitely a first, and one I enjoyed very much. Quick intros before I share some of my take-aways:</p><p><strong>Erik van Zijst</strong> <a href="http://twitter.com/erikvanzijst">(@erikvanzijst</a>) is a Principal Engineer at Atlassian, focused on Bitbucket. At 8.5 years he currently holds the title of Bitbucket’s longest contributor. <strong>James Ramsay</strong> <a href="http://twitter.com/jamesramsay">(@jamesramsay</a>) is a Product Manager at GitLab. And <strong>Briana Swift </strong>(<a href="http://twitter.com/brianamarie132">@brianamarie132</a>) is a Trainer at GitHub. <strong>CB Bailey</strong> <a href="http://twitter.com/hashpling">(@hashpling</a>), Software Developer at Bloomberg, led the panel.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*noFumVxBva4tMlJPJW9zLw.jpeg" /><figcaption>CB Bailey</figcaption></figure><p><em>CB: How did Git get such a strong foothold in OSS? </em><br>Erik: I reckon GitHub had something to do with it.<br>James: Forking is a great way to collaborate, which is not a Git feature perse but all hosts use this model now.</p><p><em>CB: What are the significant differences between OSS and enterprise workflows?</em><br>James: Branching vs Forking (respectively) is one, as well as access-control.<br>Briana: Restrictions and permissions in terms of os, language used, developer tools, are common in Enterprise environments.<br>Erik: Protocol for PRs is as well.</p><p>On scaling and the mono-repo topic, Erik says he doesn’t see a lot of sudden demand for the latter. James says GitLab does, especially from projects coming from other VCS’s. Erik: “Companies might be looking at the vendors for guidance regarding their workflow.” GitLab itseld is a monolithic Rails repo, and they built a way to coordinate things like merge requests. Erik: “Refactoring a lib that’s used across your monolith, there’s no reason fundamentally why we can’t make that work, but the tools remain unsupported. A code host would just need to put enough smarts on things like code indexes.” James agrees that a tool integrated in your Git host would be “paramount”, “that’s where the daily work takes place, you don’t want yet another tool”.</p><p><em>CB: What’s next for Git?</em><br>Briana: “Git is accessible, but it’s not easy. We can always improve how we teach it.”<br>James: “We need to get better in our visual representation, so that new users can feel confident and powerful using Git.”</p><h4>Technical contributions towards scaling for Windows</h4><p>John Briggs (<a href="http://twitter.com/jrbriggs">@jrbriggs</a>) is the Engineering Manager for the Git ecosystem team in Azure DevOps. His team is tasked with helping Git scale to support some of the largest repositories in the world with contributions to core Git, Git for Windows, and VFS for Git.</p><p>Microsoft runs on GitHub. Windows OS repo, since moved to Git in 2017, grows (massively) exponentially over time. A week in early 2018 looks like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/468/1*qldqI0_T47iBkSq8Aa0gZw.png" /><figcaption>Wait? Only 23% is actual development?</figcaption></figure><p>Available in Git 2.20 is the multi-pack-index feature John’s team has been working on. multi-pack-index is the not-so-secret sauce of Azure, which accelerates the time it takes to find an object (object lookup). Really useful when you’re in an environment with many pack-files like with VFS (Virtual File System) for Git, due to the prefetch model.</p><p>John also introduced the <a href="https://blogs.msdn.microsoft.com/devops/2018/06/25/supercharging-the-git-commit-graph/">serialized commit-graph</a>, first workshopped in the 2018 Git Merge talk “<a href="https://www.youtube.com/watch?v=oOMzi983Qmw">Making Git for Windows</a>” by Derrick Stolee and Johannes Schindelin.</p><p>For the visual learners, the current commit walk:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/738/1*sT-OUhojYdmX9GDFRUjVCw.png" /><figcaption>Reds are uninteresting, blue interesting</figcaption></figure><p>The current object walk:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/650/1*mBEUOnxqacFHlyHbHbq6Yg.png" /><figcaption>Commits, Trees and Blobs</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*qXu9E47lrm2saI6DgfWIXA.png" /><figcaption>3.1 million interesting objects (left) vs ~100 interesting objects (right)</figcaption></figure><p>And the new “sparse” object walk:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/766/1*LX6E1LMRTFI9-D19nZjgCQ.png" /></figure><p>A performance comparison (using a small topic) measured “trees parsed” (443.787 for the old algo, 37 for the new “sparse” object), git pack-objects time (10.1 seconds vs. 3.8 seconds), and git pushtime (14.9 vs. 8.5). The sparse push algorithm is currently under review in core Git.</p><p>One last thing is improving status performance by caching git status.</p><h4>How a Git based education cultivates more resilient developers</h4><p>Ben Greenberg (<a href="http://twitter.com/rabbigreenberg">@rabbigreenberg</a>) is a Developer Advocate at Nexmo. Ben is a second career developer who previously spent a decade in the fields of adult education, community organizing and non-profit management. He’s also a rabbi.</p><p>A coding education that not only incorporates Git into its education but forms its educational program around Git produces more resilient and more capable developers, or so Ben says. As a former technical coach at the Flatiron School he helped numerous emerging programmers learn the fundamentals of software development and grapple with breaking a problem down into its smallest parts through using a “Git approach”. This not only teaches good version control but cultivates a genuine trouble shooting mindset, a paradigm shift in seeking solutions. Coding Bootcamps would do good to include Git in the curriculum (and use OSS to host their materials).</p><p>The end. And then dancing might or might not have been initiated by yours truly.</p><p>*The relevant mailing list item, courtesy of CB Bailey, for reference: <a href="https://git.github.io/rev_news/2018/08/22/edition-42/">https://git.github.io/rev_news/2018/08/22/edition-42/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=50e942f06497" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Dev(Sec)Ops meetup on Kubernetes & containers, a summary]]></title>
            <link>https://medium.com/@floriendrees/dev-sec-ops-meetup-on-kubernetes-containers-a-summary-9684955d7ffb?source=rss-5fc8bd12050e------2</link>
            <guid isPermaLink="false">https://medium.com/p/9684955d7ffb</guid>
            <category><![CDATA[devsecops]]></category>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[docker]]></category>
            <category><![CDATA[devopsdays]]></category>
            <category><![CDATA[kubernetes]]></category>
            <dc:creator><![CDATA[Floor Drees]]></dc:creator>
            <pubDate>Wed, 30 Jan 2019 17:25:23 GMT</pubDate>
            <atom:updated>2019-01-30T17:25:23.038Z</atom:updated>
            <content:encoded><![CDATA[<p>Yesterday I attended the <a href="https://www.meetup.com/nl-NL/DevOpsAmsterdam/">DevOps Amsterdam</a> x <a href="https://www.meetup.com/nl-NL/DevSecOps-Netherlands/">DevSecOps Netherlands</a> meetup “Container &amp; Kubernetes Security”, at Schuberg Philis (business technology company). Below my attempt at summarizing Andreas Lambrecht’s and Andy Repton’s excellent talks.</p><h4>Security for Cloud native environments</h4><p>Andreas (<a href="http://twitter.com/vAndrucha77">@vAndrucha77</a>) is a Solution Architect at Aqua Security, securing software supply chains and the runtime of containerized and cloud native applications. Prior to joining Aqua, Andreas held consulting architect and technical solutions engineer roles at vendors like VMware, Puppet and Docker. Aqua’s team members come from places like McAfee, SAP and IBM, they partner with all the Cloud providers, and are backed by the likes of Microsoft Ventures. They literally wrote the <a href="https://info.aquasec.com/kubernetes-security">book about securing Kubernetes</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*iQOcG1L-ApjOKpYl57OkbA.jpeg" /><figcaption>Andreas Lambrecht, Solution Architect at Aqua Security</figcaption></figure><p>Shipping applications used to be a challenge. Docker changed the way we build, ship and run apps, it dramatically increased portability. But InfoSec departments often consider containers challenging to manage. Code moves too fast for risk analysis, and sometimes thousands of containers with limited visibility or control. Andreas aims to demonstrate how security controls can be continuously embedded into the application lifecycle.</p><p>Attackers are taking advantage of security blind spots, like supply chain attacks (where open source libraries are affected, like with <a href="https://blog.npmjs.org/post/180565383195/details-about-the-event-stream-incident">the NPM event-stream incident</a>) or Bitcoin miners exploiting Docker containers’ lack of runtime control, or the (several) priviledge escalations that plagued older Kubernetes versions — but more on those in Andy’s talk.</p><p>The point that Andreas is trying to drive home is that the traditional security stack isn’t Cloud native aware, and preventing breaches in large scale deployments is a real challenge. The MTTI (mean-time-to-identify) and MTTC (mean-time-to-contain) with cloud native workloads were 197 and 69 days respectively, in 2018. That’s over 2/3 of a year before a vulnerability is mitigated. Imagine the damage that can be done in this time.</p><p>Cloud native deployments lack support for effective and demonstrable security required for business-critical applications. There’s an organizational gap between Security and DevOps, that needs to be disrupted.</p><p>3 open and free source products by Aqua (in case your pockets aren’t deep enough to pay for the commercial solution), that should help you secure your infrastructure:</p><p><a href="https://github.com/aquasecurity/microscanner"><strong>Microscanner </strong></a>— A Docker image scanner that scans your builds for known vulnerabilities. Microscanner outputs HTML and JSON, and even points you to the version in which the error is fixed, and the command that will fix all your problems. Its Jenkins integration will show your logs in a pretty (or: acceptable) UI.</p><p><a href="https://github.com/aquasecurity/kube-bench"><strong>kube-bench</strong></a> — CIS benchmark scanning Kubernetes nodes against CIS checks.</p><p><a href="https://github.com/aquasecurity/kube-hunter"><strong>kube-hunter</strong></a> — k8s penetration testing, tests clusters against known attack vectors, both remote and internal.</p><p>In Cloud scale is key, and increasing the workload by a LOT in an attack scheme will make for a large AWS bill. Dev, Sec and Ops need to work together to try and mitigate similar threats.</p><p><strong>All your Kube are belong to us</strong></p><p>Andy (<a href="http://twitter.com/SethKarlo">@SethKarlo</a>), Mission Critical Engineer at Schuberg Philis, puts his finger on a sore spot: “do you think you’re in control of your container’s runtime?” In shared clusters, he argues, it is critical to ensure proper authentication, authorization and pod security best practices. Recently, CVE-2018–1002105 demonstrated the first major vulnerability in Kubernetes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*HrBclyVcFs-1875iyZJqzA.jpeg" /><figcaption>Andy Repton, Mission Critical Engineer at Schuberg Philis</figcaption></figure><p>“Today, with managed Kubernetes becoming more common, your cluster creator could be a public cloud provider, a third party company or, indeed, perhaps you. As time goes on, I believe these roles will continue to diverge as Kubernetes becomes more commoditised.”</p><p><strong>Authentication, Authorization, and Admission controllers</strong></p><p>AuthN (Authentication) and AuthZ (Authorization) are strictly seperated in Kubernetes. Different types of Authentication include client certificates, bearer tokens, basic auth, OpenID tokens, and Webhook authenticators. And with all these options, there’s really no reason not to do this right.</p><p>AuthZ defines what you’re allowed to do in the cluster, either cluster-wide, or namespace restricted. RBAC is the default authorization system today, and there are also third pary webhooks you can use for AuthZ with open source projects allowing you more complex control over your resources.</p><p>Admission Controllers intercept incoming requests and can make decisions over them. Special admission controllers are the mutating and validating webhook controllers, which allows for controllers not compiled into the API server. Recommended controllers to enable include NamespaceLifecycle, LimitRanger, and DefaultStorageClass, for the full overview see slide 7 of Andy’s presentation: <a href="https://slides.com/andyrepton/all-your-kube-are-belong-to-us#/2/7">https://slides.com/andyrepton/all-your-kube-are-belong-to-us#/2/7</a> Understanding what your cluster controllers do, is critical in not getting pwned, says Andy.</p><p>Pod Security Policies have been around for 3 or 4 major versions. As are Network Policies which allow for fine grained control over what pods are allowed to talk to what (pods) — even if Andy isn’t a fan of Network Policies.</p><p><strong>CVE-2018–1002105</strong> comes in two flavours, using the same underlying exploit. The first: an anonymous information leak via aggregated API servers, where as long as a connection is open, you could send any arbitrary (malicious) request. How to fix it? One option is to disable anonymous-auth on the kube-apiserver (the default is set to true). You could also ACL (Access Control List) your API server. But best you just upgrade.</p><p><strong>Manipulating the Kubelet for privilege escalation</strong>. Communication from the Kubelet to the API server is almost always Kubelet -&gt; API server. So when it’s not, you ought to be paying attention. If you need to exec to a pod, the apiserver needs to talk to the Kubelet, usually via port 10250. We can abuse this port to gain access if Kubelet Authentication has not been set up. Disable anonymous-auth on the Kubelet will fix this, as once again the default is true. Also consider adding authentication and authorization to your kubelets using either <a href="https://kubernetes.io/docs/reference/command-line-tools-reference/kubelet-authentication-authorization/">Certs or Tokens</a>.</p><p><strong>CVE-2018–1002105 part 2</strong>: similar to our first exploit, we are using the open upgraded connection to escalate privileges. You could disallow your users to use exec, attach, etc, but really: just upgrade ASAP!</p><p><strong>Privilege escalation via the Cloud provider</strong>. IAM in public Clouds is provided via the metadata server, so your pods appear to the metadata server as the nodes. Ensure that your nodes are created with read-only permissions to your Cloud provider. This last exploit affects Google Cloud and AWS, Andy can’t be sure about Azure.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*dszkEFdoxT3aORcMJLsquQ.jpeg" /></figure><p><em>Shout-out to the DevOps Days Amsterdam organizers, who will be doing their thing June 25 through 28. If you’re going, let me know, cause I sure will be </em><a href="https://www.devopsdays.org/events/2019-amsterdam/welcome/"><em>there</em></a><em>. Also: their CfP is open until April 1st: https://sessionize.com/dodams2019/</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9684955d7ffb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Start taking responsibility for the stuff you build.]]></title>
            <link>https://medium.com/@floriendrees/start-taking-responsibility-for-the-stuff-you-build-3e06f7d045f4?source=rss-5fc8bd12050e------2</link>
            <guid isPermaLink="false">https://medium.com/p/3e06f7d045f4</guid>
            <category><![CDATA[bias]]></category>
            <category><![CDATA[tech]]></category>
            <category><![CDATA[ethics]]></category>
            <category><![CDATA[software]]></category>
            <category><![CDATA[engineering-ethics]]></category>
            <dc:creator><![CDATA[Floor Drees]]></dc:creator>
            <pubDate>Wed, 19 Sep 2018 17:29:02 GMT</pubDate>
            <atom:updated>2018-09-19T17:29:02.798Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Below is the transcript of my CodeDaze 2018 talk.</em></p><p>The <a href="https://insights.stackoverflow.com/survey/2018/">Stack Overflow 2018 survey</a> of 100,000 developers provided a ton of insights. Some things remained more or less the same, compared to the outcomes of 2017 and 2016. Major difference: for the first time Stack Overflow included questions about ethics.</p><p>Answering the question who is primarily responsible for considering the ramifications of AI, 47.8% of the respondents answer ‘the developers or people creating the AI’, 27.9% answers ‘A governmental or other regulatory body’, 16.6% says ‘prominent industry leaders’ and 7.7% says ‘nobody’.</p><p>Whereas 58.5% of the respondents claim they’d never write code for an unethical purpose (define unethical purpose), 36.6% built in a safety net, answering ‘depends what it is’, whilst the remaining 4.8% says they would. I wouldn’t be surprised if the subjective frame of this question reeled some politically-correct answers here, but even if it hasn’t, 40-ish % would probably ‘do evil’.</p><p>Respondents overwhelmingly claim that the blame for unethical code rests on ‘management’ (57.5%), or the person who came up with the idea (22.8%), versus the developer who wrote it (19.7%).</p><h4>Who are we? What do we want?</h4><p>Software makers yield enormous power and yet are still a homogeneous group. Our creations are inherently bias, exclusive and flawed. We have a superpower: we can make computers do things. We build tools, and ultimately responsibility lies with us to think through how those tools will be used.</p><p>Let’s talk about the responsibility and accountability we individual builders-of-the-web should claim. Using our exceptional privilege of mostly white, mostly male , sought-after tech workers push for a change.</p><ol><li><strong>Push for products and services safe for your users, present and future.</strong></li></ol><p>Underrepresented groups are unequally affected by products that are not strictly safe. 1.1 Companies based in the EU are under some pressure to be GDPR compliant. Certainly that pressure could be intensified, were local authorities not so understaffed and uncertain about the way to proceed, dealing with a law that was intentionally kept vague to survive the next decade of technological ‘disruptions’. 1.2. Except for bad data collecting practices, I’ll dive into products that are deliberately addictive or otherwise harming. 1.3 And security.</p><p><strong>2. Push for fair working conditions for all.</strong></p><p>As an employee of Tesla, Uber and the likes you also hold responsibility over how its warehouse workers, its contractors, gig-economy dependents are treated (and discouraged from unionizing).</p><p><strong>3. Push for inclusion of underrepresented groups, on every level.</strong></p><p>3.1 In the workforce, and not just for optics. We need to retain these people and make sure they have the same opportunity to advance. <br>3.1.2 We can refer people, as long as your network doesn’t comprise of white men only. But refer responsibly, if you know the people you bring in won’t have the opportunities they need and deserve, don’t. <br>3.1.3 Dedicate your time to coaching/teaching others, better yet have your management let you do it in company time.</p><p>3.2 But also inclusion in the audience you build solutions for.<br>3.2.1 Push for diverse teams, that uncover the blind spots we have. <br>3.2.2 Push for code review standards, Colin’s talk can be a helpful reference. 3.2.3 Push for adopting a CoC for anything your company releases open source, for the local meetups your company sponsors, or conferences you have a booth at.<br>3.2.4 push for diverse test panels and focus groups.<br>3.2.5 Push for accessible products, which you should know how, after Adrian’s talk yesterday.</p><p>Let me get into these three items one by one.</p><h3>Pushing for safe products and services</h3><p>The modern approach to data protection <a href="https://medium.com/s/trustissues/user-agreements-are-betraying-you-19db7135441f">relies roughly on the idea</a> that if users were given notice of a company’s data practices and choose to continue using the service, then they consented to those practices. Transferring many of the risks of data processing to users, while maintaining the status quo, is convenient but not fair. To this day, these agreements largely exist to not inform users in an intelligible way.</p><p>Consumers more often that not trade data for services without grasping how their information can be stored, shared, used, or even how much their information is worth. Agreements are notoriously vague and present users with take-it-or-leave-it offers that deprive consumers of choice. Real transparency means that the user is fully informed without having to wade through pages full of legalese. Nor by <a href="https://medium.com/s/trustissues/facebook-fabricates-trust-through-fake-intimacy-b381e60d32f9">silencing the individuals on your freshly erected board overseeing ethics with an NDA</a>.</p><p>The illusion of agency via terms and settings to keep the data machine running is nothing but a lever. We need to create rules that don’t require or even expect users to read these agreements. We need to make our companies trustworthy regardless of the control we are given. We need to make reasonable care the default. It’s not the users responsibility to police the industry.</p><p>Defaults matter. I’m going to guess that 90 percent of users don’t change the default settings of a technology product they buy or use within the first six months. Let alone thereafter. In most cases, companies don’t need all that data to provide their services. So what if we flipped the defaults? There’s the approach of lean data practices. Makes for less data to lay out on the street in case of a breach.</p><h4><a href="https://t.co/wgBUaLHiaG">What’s good for capturing human attention is bad for humans</a></h4><p>The dopamine hits we get from likes on our Instagram photos, gamification elements that do more harm than good. Hooks, those little red dots of notifications piling up. Depression levels are surging, no doubt in some ways fuelled by the Fear of Missing Out.</p><p><a href="https://medium.com/@azumbrunnen/the-rise-of-anti-notifications-6c92e355612b">Few inventions affected our relationship with technology as much as push notifications</a>. Whereas before most of us felt in charge to decide when we want to interact with technology, it is technology that is now largely making that decision for us. Whereas we used to get notified about things that were all about us, we now get notified when random friends like each other’s pumpkin spice latte.</p><p>A litany of technologies, like the internet, have contributed immensely to human welfare. Such technologies are worth preserving with rules that mitigate harm but accept reasonable levels of risk. Facial recognition systems however are not among these technologies. Consent rules, procedural requirements, and boilerplate contracts are no match for its risk of exploitation.</p><p>The American Civil Liberties Union (ACLU) and other coalition partners <a href="https://medium.com/s/story/amazon-needs-to-stop-providing-facial-recognition-tech-for-the-government-795741a016a6">urged Amazon to stop providing government agencies with facial recognition technology</a> and services associated with its Rekognition system. The system poses a general problem (no anonymity in public) with unevenly distributed consequence (that are more threatening to minorities and other vulnerable groups), and thus a major threat to civil liberties.</p><p>Amazon’s response suggested that if problems ever do arise, appropriate policy correctives should be followed, placing the responsibility with lawmakers. But the irresistible power of a system like Rekognition poses a severe threat in the hands of law enforcement. In a test the ACLU recently conducted of “Rekognition,” <a href="https://www.aclu.org/blog/privacy-technology/surveillance-technologies/amazons-face-recognition-falsely-matched-28?">the software incorrectly matched 28 members of Congress</a>, identifying them as other people who have been arrested for a crime. Which is why Amazon employees, shareholders, a coalition of nearly 70 civil rights groups, over 400 members of the academic community, and more than 150,000 members of the public have spoken up to demand that Amazon stop providing face surveillance to the government.</p><p>Sometimes I think the best course of action might be to avoid the technology all-together, before its use will become the new normal.</p><p>Google Duplex <a href="https://techcrunch.com/2018/05/10/duplex-shows-google-failing-at-ethical-and-creative-ai-design/">deployed human-sounding vocal cues for its bot</a>, such as ‘ums’ and ‘ahs’ — to make its “conversational experience more comfortable”. I don’t think you need to deceive someone to give them a better user experience.<br>Google let go of it’s don’t be evil slogan as early as 2014, and in my opinion Duplex is just one of the missteps against it’s old claim.</p><p><a href="https://www.bloomberg.com/news/articles/2018-06-07/google-renounces-ai-for-weapons-but-will-still-sell-to-military?via=indexdotco">Staff protests forced Google to retreat from the Maven contract</a> in June. Yet Google’s cloud-computing unit, where the company is investing heavily, wants to work with the government and the Department of Defense because they are spending billions of dollars on its services.</p><p>In <a href="https://blog.google/topics/ai/ai-principles/">Google’s newly penned up AI principles</a>, the company pledges not to do anything in violation of accepted human rights laws. In a press release it went on to say it will integrate the principles into existing product-review processes and plans to set up an internal review board to enforce the guidelines.</p><p>Then, what about the <a href="https://medium.com/@doctorheadcrash/we-have-a-killer-robot-problem-but-not-the-one-you-think-2afc538cd2d6">massive, systematic risk posed by our poorly secured home automation devices</a> we’ve indiscriminately connected to the internet? Most modern robots are periodically connected which exposes them to a wide range of threats. There is a large and growing installed base of legacy devices with easily exploitable security flaws, in hospitals and in our homes.</p><p>How can we expect inexpensive consumer IoT products to be properly secured when we seem to have difficulty protecting voting systems? There are no widely accepted security standards for IoT devices, and security engineering guidelines are still in the proposal state.</p><h4>There is no such thing as self-regulation</h4><p>In 2001, Meg Whitman, then eBay’s chief executive, described the company as a “dynamic self-regulating economy.” When eBay has taken steps to address problems in its economy, it has tended to do so by providing <a href="https://medium.com/new-york-times-magazine/want-to-understand-what-ails-the-modern-internet-look-at-ebay-89eeb98c6d5b">new systems to encourage users to work things out themselves</a>. To deal with sellers who ship slowly or deceive, eBay lets users rate one another. In theory, this starves bad actors out of the marketplace — a poor karma will prevent a seller from selling. In theory.</p><p>An Uber driver might be dinged if a rider doesn’t like their personality or choice of music. Facebook tried ranking publishers by trustworthiness, based on feedback from users. Upon discovering a problem, a platform company’s first instinct is to find a way to expose it to the ‘wisdom’ of the crowd. eBay users’ participation in the site’s governance is at its core, snitching on one another.</p><p>Normally we think that organizations with a lot of power have an obligation to use that power responsibly. But the leaders of the largest technology companies <a href="https://www.vox.com/new-money/2016/11/16/13637310/facebook-fake-news-explained">have by and large resisted thinking of themselves in those terms</a>. They like to think of their sites as neutral platforms that help users share information and connect.</p><h4>Facebook the Internet not is</h4><p>Facebook’s Internet.org project, bringing internet connectivity to countries underserved, was banned in India for violating net neutrality. Yet, thanks to heavy subsidies that <a href="https://medium.com/new-york-magazine/facebook-used-the-philippines-to-test-free-internet-then-a-dictator-was-elected-a8486f79d4ad">keep Facebook free to use on mobile phones</a>, Facebook has completely saturated countries where it pioneered, like the Filippines. Because using other data is expensive, for most Filipinos the only way online is through Facebook.</p><p>Roughly 69 million Filipinos have access to the internet today, and 97 percent of them have a Facebook account. I don’t need to remind you that Facebook’s News Feed algorithm promotes anything that attracts attention, amplifying bad actors.</p><h3>Fair working conditions for all</h3><p>The funny thing about <a href="https://medium.com/the-guardian/the-undertakers-of-silicon-valley-how-failure-became-big-business-b8d81d8f6002">the tech industry’s narrative about itself</a>, is that failure was always assumed to be temporary (bugs can be fixed); where for everyone else, it’s devastating. Taxicab companies going out of business because they’re losing money is considered great creative disruption. This double standard justifies the exploitation of Uber drivers hardly making a living.<br><a href="https://www.wired.com/story/amazon-mechanical-turk-bot-panic/">Mechanical Turk is bot-ridden</a> largely because of the unfair compensation, yet we rely on it for (social science) research. That doesn’t seem right.</p><p>Healthy working conditions too, where everyone can largely participate by default. Last year I interviewed with a FinTech startup, which name starts with an <em>R</em> and ends with <em>Evolut</em>, and my 2 consecutive interviewers told me 100h workweeks were no abnormality and challenged me that I couldn’t keep up being a mom, which is prolly true, and I wouldn’t want to, not as a mom and not as a human being, but that’s besides the fact that they shouldn’t have this information in the first place. Also: equal parental leave.</p><p>One of DHH’s <a href="https://twitter.com/dhh/status/1031185841295523840">most recent Twitter storms</a> revolved around a company promoting crazy hours in their job ad.</p><h3>Inclusion of underrepresented groups</h3><h4>Bad inputs lead to bad outputs</h4><p>Because machine learning technology is deployed by governments and companies in fields ranging from criminal justice to recruitment, workforce biases in the tech sector are amplified across other industries and the public sector. AI algorithms learn from historic data sets, so they are more likely to hire in the image of previous staff instead of helping to tackle unfair under-representation.</p><p>Big tech, and specifically social networks are mostly limited by a workforce that doesn’t understand what abuse looks and feels like. Equal opportunity today looks like this: the internet is the modern front line of feminism, amplifying the voices of (predominantly white) women around the world, as much as it provides space for so-called ‘Men’s Rights Activists’ to form alliances.</p><p>The ramifications online affect different people in different ways, precisely because of the inequity of the society it mirrors. It affects men and women differently. But more-over it affects people of colour differently. It affects trans and gender fluid people differently. It affects people differently on a whole spectrum.</p><p>A handful of tech giants have become the self-proclaimed guardians of global speech, amplifying certain kinds of voices and limiting others, referring to their often opaque and shifting standards. Although these companies — Facebook, Google and Twitter above all — are now integral to just about every corner of our lives, yet face little regulatory oversight.</p><p><a href="https://medium.com/the-atlantic/the-most-powerful-publishers-in-the-world-dont-give-a-damn-2fe44f4538ae">Facebook is the largest media company in the world</a> and Zuckerberg the most powerful publisher. Twitter is a smaller media company, yes, but still wildly influential in many ways, not least of all because it is known for being the platform of choice among celebrities, politicians, and journalists. Also: conspiracy theorists.</p><p>Twitter founder Jack Dorsey rejects the idea that publishers are responsible for the quality of what appears on their platforms. Instead we should know to rely on other publishers to uphold standards such as seeking the truth.<br>Systems users rely on are ridden with hidden, pervasive and often unintended bias — the sort that led Google to return pictures of black people for searches of “gorilla.” Google is made by humans who have preferences, opinions and blind spots and who <a href="https://medium.com/the-new-york-times/heres-the-conversation-we-really-need-to-have-about-bias-at-google-ca6cf00f63b3">work within a corporate structure that has clear financial and political goals</a>. Not to say that not for profit or open source organisations are inheritely inclusive, as Coraline pointed out in her EuRuKo talk The Broken Promise of Open Source, last month.</p><p>Because Google’s systems are increasingly created by artificial intelligence tools that learn from real-world data, there’s a growing possibility that it will amplify the many biases found in society. And it’s not just search. Google owns YouTube and Gmail, and it makes the Android operating system and its app store. It is the world’s dominant internet advertising company, and through that business, it also shapes the market for digital news. Google’s algorithms favor recency and activity, which makes it vulnerable to being manipulated in favor of misinformation.</p><p>Diverse teams are more likely to flag problems that could have negative social consequences before a product has been launched. Diversity campaigns at companies such as Google have failed to significantly shift the predominance of white cis het and Asian men in their technical workforces.</p><h3>So what are we to do?</h3><p>Ask why. A lot. Why do we need this feature, this product? Why do we need to track users every move? Why do we need another white male hire? Another white male consulting team? Why is this person not involved in this meeting? Encourage your team to follow your example. More than <a href="https://t.co/qFPaKpVcrQ">4,000 employees signed a petition calling for the cancellation of Google’s Project Maven contract</a>, some even resigned over the topic. We are with that 4,000.</p><p>Don’t be the douche hiding behind the “I am just the developer” line. Being a good software engineer does not qualify you to engineer society, politics, economics, and beyond. But that’s exactly what we’re doing, every day. As tech companies take on increasingly vital roles in global commerce and culture, we have the power to shape the way they operate and the ethical standards they abide to. Tech workers are the only point of leverage because we’re no commodity (until we’re replaced by robots and have to survive on UBI).</p><h4>When you invent the ship, you also invent the shipwreck</h4><p>The famous Paul Virilio quote above applies well to the effort Apple and Google claim where they announce features to help people cut back on their tech use.. Even Instagram has recognized the “zombie scrolling” phenomenon and rolled out a new interface to help users break the habit.</p><p>While at first glance, it appears <a href="https://medium.com/behavior-design/the-real-reason-apple-and-google-want-you-to-use-your-phone-less-37b2b7f1384">their business model would benefit from addiction</a> — the more you use your phone, the more money they make through the apps you buy and the ads you view — for their long term goals however, Apple and Google don’t want you to get addicted. Addiction is a compulsive and harmful behavior. And they rather see you form healthy habits with your digital devices.</p><p>Incorporating safety features, their behaviour mimicks that of car makers introducing seat belts in the 1950s.</p><p><a href="https://medium.com/bloomberg-view/no-twitter-healthy-conversation-cant-be-engineered-7f94325770ef">Facebook and Twitter are attempting to curb toxic behaviour on their platforms</a>. If only they wouldn’t solely rely on the worst technology to do so, AI. Healthy discourse can’t be engineered, even if it’s compelling to try. Until there are real consequences to insults, harassment and deception, the conversation cannot be healthy.</p><p>Google wants to cut fake news sites off from access to its vast advertising network, depriving them of a key revenue source. Facebook quickly followed suit with its own ad network. Social media has drastically lowered barriers to entry in the news business. Twitter, Google, and especially Facebook have gained a huge and growing influence over what news people see. It only makes sense that they fix the problem.</p><p>There have been proponents of <a href="http://](https://medium.com/s/story/please-jack-kill-anonymity-on-twitter-60ff79bf8ec4)">ending anonymity on Twitter</a>. I wonder if the trade-off would be worth it? Could Arab Spring have taken place on a platform that needs verification for you to participate? Anonymity is necessary, but what if not you perse, but someone or something (some entity) could lend you verification? An interesting discussion to have.</p><p>I’d urge you not to take requests at face value. Technology is moving so fast. There is no way laws can ever keep up. We only seem to ask ourselves “Can we do this?” “Sure, we’ll just move fast and break stuff.” And we never stop and ask “Should we do this?”</p><p>How can organisations working at the bleeding edge of technology design for and invest in process that informs governance and operationalises ethics when the critical analysis of unintended effects largely takes place outside of the locus of invention. Today the process of inventing is not influenced by critical analysis.</p><p>The author of the article ‘<a href="https://medium.com/@alixtrot/working-ethically-at-speed-4534358e7eed">Working ethically at speed</a>’ proposes a setup of agile ethics, borrowing from the agile development method. A process whereby ideas are researched, prototyped, and then shaped by frequent testing with a potential user group, preventing harmful creations from being flung into the world before their potential impacts are assessed. When non-diverse teams use agile methods they often have ethical blind spots (wilful and not) that can embed harm in what they build. Agile ethics is a process to operationalise values, identify and address ethical challenges of our innovations, and iteratively refine once you launched.</p><p>Hayley’s talk about licensure yesterday rang true to me. As a society, we recognize that when consumers dine out, they may be putting their health at risk. <a href="https://t.co/8w2AbvXvCy">Public information about interacting in the digital world is lacking</a>, which is striking given the relative severity of the threat. Many disciplines have to take some equivalent of the Hippocratic oath. But not us. Why?</p><p>The First Amendment does not apply to private companies — but these companies are so powerful that their decisions do have political consequences. PayPal began to prohibit donations that “promote hate, violence and intolerance.” GoDaddy and Google refused to host a neo-Nazi website.</p><p><a href="https://medium.com/thewashingtonpost/feminists-were-right-ignoring-online-misogyny-has-deadly-consequences-e7667c7aa4ce">Giphy was recently booted by Snapchat and Instagram</a> after an explicitly racist GIF outraged users — and Giphy responded by cleaning out its database, signifying a shift in how we hold social media platforms accountable. <a href="https://medium.com/@WIRED/the-guerrilla-wikipedia-editors-who-combat-conspiracy-theories-15c29b5f99c1">Volunteers step up to combat conspiracy theorists on Wikipedia</a>. Not all hope, it seems, is lost.</p><p>The effects of decisions or actions based on AI are often the result of countless interactions among many actors, including designers, developers, users, software, and hardware. This is known as distributed agency. <a href="http://science.sciencemag.org/content/361/6404/751.full">With distributed agency comes distributed responsibility</a>.</p><p>Existing ethical frameworks like AI4People and IEEE, the European Union (EU) strategy for AI, the EU Declaration of Cooperation on Artificial Intelligence, and the Partnership on Artificial Intelligence to Benefit People and Society are so important, yet largely unknown to most of us in this room. Let’s change that.</p><p>Thank you.<br>Also, please watch Thompson’s chilling talk ‘<a href="https://www.calebthompson.io/talks/dont-get-distracted/">Don’t get Distracted</a>’.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3e06f7d045f4" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>