Audio Chat, Licensing, Docker, 32bits, NCSC-FI, H3, Python

1. Async (group) audio chat - I've asked this from many chat app projects. I don't know if it's useful, but most projects seems to think that it's not, or that it's too hard to implement. The question: Why there's no option for async automatic audio play? - It would make quasi realtime group audio chat viable. Great for group communications when writing isn't an option, nor constantly on going group call is necessary. Technically it should be trivial, but still it seems for many app developers find it too complex concept to implement. All that it takes is: "autoplay new received audio messages" and that's all. It's like a hot line without keeping the hotline hot all the time.

2. Checked out a few very permissive licensing models: Public-domain-equivalent license, 0BSD and Unlicense (all links @ Wikipedia).

3. Some docker images are annoyingly badly outdated. Because the services provided by those images are quite simple. I've replaced those mostly with bubblewrap (@ wiki.archlinux.org) or using Ubuntu base Docker (@ Wikipedia) image, and then just mounting startup scripts and software from read only bind to the image and running it. This allows my to trivially update of software and image separately. Just copy new files to the host directory and restart container, done. Sure I could do my own docker images as well, but that just adds extra friction, steps and overhead to the simple process.

4. Daily WTF - Got a serious question of one server application is tested and runs well with 32-bit Windows Server. - Uh, it probably does. But who uses such servers in production anymore?

5. One project was a bit slowish and consumed a lot of bandwidth. After analyzing the causes, there was one simple thing to be done. The application transferred huge amounts of bloated JSON data without using any data compression. After I added gzip support to it, now the overall process runs about 70% faster. - Nice! And of course we also save a lot of bandwidth.

6. Watched latest - The National Cyber Security Centre Finland's (@ kyberturvallisuuskeskus.fi, their official site) Supply Chain Security lectures.

7. HTTP/3 (@ blog.apnic.net) - Very nice post by APNIC about HTTP/3 (@ Wikipedia) and why it's necessary. Yet as expected, this is one of key internet technologies, and I've been posting about this topic for several years. Nothing much new in the article. But if someone doesn't know what HTTP/3 means, then it's a great time getting to know it. Some network gear could suffer from UDP performance issue, if lot of traffic is shifted from TCP to UDP.

8. Python 3.12.0 (@ python.org) - f-string improvements are awesome. per-interpreter GIL... Whoaa-aa, the GIL has been discussed like, well, forever. My code structure is currently using multiprocessing in cases where I need multiple GILs. But it's great that part is now getting test driven and probably included in next Python 3.13 version as Python standard library feature. Comprehension inlining, more speed, way cool. Improved Error Messages, totally awesome. Especially for people starting to program, old Python error messages were often unfortunately just enragingly confusing. asyncio socket writing performance increases using sendmsg are also great. os.listdrives listvolumes listmounts, something I've been missing several times. path.walk makes traversing trees easier. No need to reinvent the wheel. Temp file option, delete_on_close, nice. avoids with .... finally structure making it cleaner.

9. NYM anonymity network roadmap got updated (@ nym.tech.net) - It's been really slow progress, interesting to see if anything ever comes out of it.

10. UUIDv7 (@ buldkite.com) - A nice way of having chronologically sortable UUIDs. Also see: RFC draft (@ datatracker.ietf.org)

11. Too much discussion about databases and performance. And way too much speculation. Why they lack experience, measurements and factual data?! I can't take this anymore. Had to write them several sample programs demonstrating the simple concepts and things they didn't get. That worked, when they were able to see the code and test run it them selves. Instead of having theoretical discussion with invalid expectations and reasoning. After all it took just four hours to make it clear, that there are better ways of doing things than how they're doing those now. - GG

2024-08-18