nate berkopec @nateberkopec@mastodon.social

What's the number one thing you would like to see me make/publish in the next 3 months?

Pushed a new app last week and can confirm that Heroku is just as easy to use as ever.

MRI RubyのGlobal VM Lock（GVL）は、何かと説明が的外れだったり誤解されたり批判にさらされたりしがちです。My article in 日本語, translation by @hachi8833@twitter.com https://techracho.bpsinc.jp/hachi8833/2020_05_27/92042

Anyone know why Webpacker defaults to a single JS pack in development but separate JS/CSS files in production? Probably the only thing I haven't liked about it so far.

Anyone out there using GnuCash on a regular basis? Any tips?

RT @TokyoTechMeetup@twitter.com

Tune in on Tuesday, June 2nd for a lightning talk by @nateberkopec@twitter.com on building out mental models for web application performance and scaling https://www.meetup.com/tokyotechmeetup/events/270419853/

🐦🔗: https://twitter.com/TokyoTechMeetup/status/1265812626585513985

Finished "The Goal". It's interesting, basically Lean production but told in a narrative format for a Western audience. A bit basic on the details + the cultural references haven't aged well. Shorter + more useful is Taiichi Ohno's "Toyota Production System".

It's really important to know the ratio between I/O and CPU time on your Ruby web apps + background jobs. If you know that ratio, you can provision process/thread counts correctly. W/o it, stabbing in the dark most of the time.

RT @maciejmensfeld@twitter.com

🚀🚀🚀 Malicious packages in @rubygems@twitter.com? Not anymore! I've built and OSS supply chain security platform for Ruby: https://diffend.io - it acts as a free and public proxy for RubyGems blocking any malicious gems from even being downloaded! MUCH more to come!

🐦🔗: https://twitter.com/maciejmensfeld/status/1264939229307691009

maybe you could reify providers, then reify "lang stacks" which just represent the ~5 bash commands you need to execute to get a language running on a fresh ubuntu box. so the final command would be like $ my-command up --lang=ruby --provider=do

it probably needs 3-4 more hours of polish and it would actually be a pretty useful set of bash scripts for standing up remote dev boxes I think.

I made a bash script that glues Terraform, Digital Ocean and Mutagen together to give you a pop-up remote Ruby development environment w/port forwarding/filesync from the current directory: https://gist.github.com/nateberkopec/5e1d81c8b5f4865c19e18ecae73c77ae Using it to stand up temp Linux dev boxes for Puma work

Re: remote work arbitrage, its a fantasy that the entire industry will suddenly start paying us all SF salaries. The bold move would be to pay everyone the same, but pay under-market w re: to SF but above-market re: to US average.

Yikes, if you're using Docker for Ruby apps in non-public environments (i.e. in dev), looks like there's a security setting you can change to make your app up to 2x faster https://mamememo.blogspot.com/2020/05/cpu-intensive-rubypython-code-runs.html

The more I look into deployment tech, the more I think Hashicorp had all of this shit figured out about 5 years ago.

TIL @brendangregg@twitter.com wrote an entire collection of several dozen troll system performance tools: http://www.brendangregg.com/specials.html

This expresses a feeling I've had for a while: https://lobste.rs/s/dyeuqm/i_want_own_database_my_apps_use I want SAP but for personal data. Everyone uses the same DB, and I control it. Maybe impossible. Maybe too complicated. But feels, at least, _possible_.

英語を練習したい場合はDMしてください。みなさんとともだちになりたいです. 日本語が上手じゃありません、でも英語がじょうずです

There is no perfect browser load metric, because what we really care about cannot be measured by a browser: the moment a user "feels" a page is loaded and starts to do what they requested (read the page, fill out the login form, etc). Depends human perception + the task at hand.

Rules of thumb for web application scales: 90% of apps are doing 1 to 1000 requests per second. The Big Shops (the name-brand sites with wide audiences) are doing 1000-100k req/sec. Your blue chip shops (Alexa worldwide Top 100) are doing 100k-1mil+ req/sec.