2D codes, Submarine Cables, Static Web, Btrfs

1. Baltic cable cuts - Some routes are now quite curious. A few mirrors I use in Latvia are now being routed Helsinki, Stockholm, St.Petersburg, Moscow, Riga. Quite a rout. Also latency bumped up from 7 ms to 31 -  62 ms with large jitter. Everything is of course still working, but performance to that specific location with their peering and routing is now clearly worse than it was earlier. It's also sometimes interesting when people claim that the time of the cut isn't known, even if it's very clear from different monitoring platforms immediately.

2. Again the age old stuff is in news: GS1 DataMatrix, GS1 QR Code, and GS1 DotCode - Yeah, nothing new about 2D codes (@ Wikipedia). Maybe those will be in consumer use in future, maybe not. Some products already have QR Code's for consumer information. Usage of GS1 keys allows encoding different agreed data types in the code. Yet I haven't yet seen any real world use cases for the the DotCode.

3. Qwen Coder - I use it too. It's nice. I've fixed many small issues I knew existed, but just didn't bother to fix or improve. Now it was easy and quick enough. - Productivity improvement. I've also recommended Qwen Coder for several friends, especially those whom have already some challenges getting the stuff implemented correctly in the very first place. - I'ts good. - When the AI race is on, we're going to see much better versions in future. - Of course it won't solve the classic problems of ambiguous specifications. But at least as classic examples of lacking exception handling is generally quite nicely handled.

4. I've kept a few quite complex yet light (small amount of complex code) projects as a personal challenge and my very own little of a coding “Sudoku”. Yet now I'm ready to let AI to check the issues on the code I know already about. If it can't directly address the bugs, at least it can restructure and refactor the code in a way, where finding and fixing the bugs should be much simpler and easier. - Give the Qwen Coder a try. Even if it wouldn't create perfect code it's a good sparring partner or coaching tool. - Yet this field is moving very fast right now.

5. Uh, Second Windows Server 2025 having strange problems with IPv6. Again, after contacting the support the issue got fixed, and they referred to IPv6 FAQ which of course doesn't contain anything new to me. I didn't change any settings and it started to work. - Highly confusing. This time they didn't say anything else than that I should check the settings. - Maybe there's a problem with Windows Server 2025 Datacenter deployment or maybe not. Who knows. At this point I've got strange feeling, that I'm going to see this issue again, and I'm still clueless what it actually is.

6. I found out about Static Web Server (@ static-web-server.net) and it's support for Zstd encoding. No, I haven't used it yet, and I don't have any use cases for it right now. But I'll keep it in mind, if I'll ever need it. Yet, because I'm very familiar with Nginx, I usually prefer using it.

7. Bypass Go SLL - Started to use their certificates, those are valid for 180 days, much nicer renewal interval than with LetsEncrypt's 90 days. Nice, for cases where the environment is so closed that any automatic renewal isn't basically possible. - Update: On the posting date, and they announced shutting it down!

8. I did some networking protocol analysis with AI backend and found out funny event. The final message before closing the session was: "event: done", "data: I'm Mr. Meeseeks. Look at me." - Lol!

9. Btrfs snapshots, I've got a few external drives, which data is automatically batch updated. I added to the batch automatic snapshot creation / deletion steps, so that it creates snapshot before update and deletes previous snapshot. Using this methodology allows me to easily return to the previous snapshot, in case something goes catastrophically wrong during the update. - Nah, it shouldn't go wrong, everything always works, they say. - But just in case. Using snapshot in that case has minimal performance impact / cost, so why not?

SMR / BTRFS / SQLite3 performance 

I know this is sick and twisted. That's exactrly just why I did it. And if there would be race conditions, this would probably show those as well.

Initial setup: 64 parallel threads - 1024 increments to the shared counter in the database each in individual transaction per thread.

After execution checks: Shared counter in RAM: 65536, Counter in database: 65536, Counts match, results are valid.

SMR + btrfs + SQLite3 -> Task executed in 98.38 minutes.

Fast SSD -> Task executed in in 423.82 milliseconds.

Haha, quite a difference. But the whole main point is that the counts do match and nothing gets broken.

2025-08-24