Weeknotes 2022-05-16

Trying to export a HDR video from my phone led to unexpected results. The colors were really washed out and the video was slightly overexposured. Using HandBrake I finally got a good looking mp4, but then someone uploaded this file to dropbox, and it looked bad for people downloading it from there. Strange.

Docker Desktop on Mac just broke during update and I couldn't restart it. Had to re-download it. Wow, this is some next level brokenness.

My PR to fix a bug in jupyterlab was merged. I couldn't find a solution to the "you cant mock a function which you call from another function in the same module" problem. Instead Frédéric Collonval circumvented the issue by using a test helper that dismisses the openen dialog instead of trying to mock it, thanks a lot.

Articles

Software

Videos

Podcasts

  • #282 – David Buss: Sex, Dating, Relationships, and Sex Differences Lex Fridman Podcast

    Some nice anecdotes

  • Über Können, Nichtkönnen und den Weg dazwischen Was denkst du denn?

  • Revision 529: Richtig schätzen Working Draft

    Hmm, there are quite some areas where I would respectfully disagree. For example: This notion of avoiding to spend too much time on optimizing your code. In reviews I rarely see code that is "too optimal" or has "too few bugs". In my experience people often say things like that when they mean to say: Optimizing this code is just not worth the effort. Which is a business decision. And if programmers make those decisions without telling a soul they make it impossible for the business to adapt (yes, this is bad). The same problem arises if you optimize for being correct on your estimations (good for you) by just doing small or simple stuff (bad for business).  Of course, when your customer is never ever talking to you - maybe because it's a pharao resting peacefully in a pyramid - it might pay off to cut some corners 😉.

  • Warum Java für die Cloud so gut ist, mit Adam Bien programmier.bar

    Rather funny episode starting with the question: Why is Java so popular? Making a really long lived entity like a programming language (if successful) depend on a rather ephemeral entity like a company never resonated with me. Even before oracle hit the fan. But good marketing can do wonders. And I never got this whole serverless thing. The best explanation I saw is: serverless means pay by usage. Which might appeal to the needs of people coming from the host world, because hosts are so fucking expensive you have to be able to assign the blame. But there's a reason why hosts are dead. Personally I would prefer to make a service cheap enough you dont have to care instead of paying magnitudes more to be able to write more detailed bills. And there's always on premise, which I like too. But kubernetes and rancher being "perfect for on premise"? Wow, I think there's another marketing wonder in progress.

  • WR1362 Ernergiespeicher, Schwarze Löcher und Sex WRINT Wissenschaf

  • Episode 1 - Origin Story Sad Python Girls Club

    New Python Podcast 🤩, lets see how it goes..

Papers

  • Note on Distributed Computing

    Someone posted this on a twitter thread discussing why HATEOAS is a newer and more innovative concept than all this "modern" SPA solution being basically good old client/server/RPC architecture which attempted to compete with the web several times (CORBA, SOAP, etc), but failed each time.

Return to blog