Croc, AI, AVIF, Scripting, Ubuntu, SimpleX

1. Croc (@ GitHub) annoyance, why it's so badly implemented. And why IP parameter doesn't do DNS resolution and so on. Multiple over lapping issues, which making usage slow and agonizing. Well, you can get it to work, when you thinker and test every possible alternate configuration. But uh, sigh... Also transfers are really annoyingly slow without doing all this tinkering and configuration. Local relay mode doesn't seem to be working. It's not documented. But can be worked around. The best method I found out after hours of testing, was to run additional local relay which is kind of totally silly, but it was the best possible configuration. First I start local relay with password on the sending or receiving server, and then tell both clients to use it. - Ok. The port configuration is also totally messy compared to traditional platforms. Even if port is specified the configuration isn't being followed directly etc. Why files are hashed in advance? It makes starting the sharing very slow and requires extra firewall configuration which is time confusing and even in some cases expensive. Also if you don't specify that extra password for relay then it'll be global open relay... Nothing special. But some options could be so much more straight forward. I think I'll need to find program just like Croc, which does exactly the same, but it is implemented in way more user friendly way.  As example using parameters --host --remote and --port without needing additional relay processes and overly complicating simple things. From the slogan: "Easily and securely send things from one computer to another". I wouldn't agree with the easily part. I expected straightforward experience, but it took hours and was very messy. Configuring sftp server would have been faster than this. But it's good learning process. Next time if and when necessary, I'll do it so, that first I launch the relay and the configure clients to use it. Without even trying direct transfer. - Quickly checked out wormhole-william (@ GitHub) which turned out to be even worse.

2. Tested many AI (@ Wikipedia) models, Gemini 1.5 Pro, Llama 3.2 90B, o1 Preview, GPT-4o. So far I really like the Gemini, it's fast. But after having a chat, if I'm unsure about something, I ask the o1 preview to check whole thread. Then it points out many of the mistakes made by the other models. - These are moving so fast that info in my blog is seriously outdated!

3. So many programs have more or less broken AVIF (@ Wikipedia) handling.

4. Element X got totally broken AVIF handling. Well, I guess they'll manage to fix it at some point. But in general it's quite annoying that image compression formats create so big differences, even if these all should be similar bitmaps to show after decompression.

5. Fixed several fatal flaws in bash and fish shell scripts. It's so insanely common that escaping and statements aren't exactly correctly formed, or if statements and scope grouping is invalid etc. And as a word of warning, many AIs got Fish Shell scripts completely wrong, do not trust any of the code given. Yet this shouldn't be any kind of surprise to anyone. It's also kind of funny how insanely bad Copilot (@ Wikipedia) is with PowerShell scripts. Yet another reason why I like writing control flow in Python and just calling a specific external commands when necessary.

6. Finally updated last of my Ubuntu (@ Wikipedia) 22.04 LTS systems to 24.04, now every instance has been updated to latest release. Yet some of the systems will be shut down in near future.

7. SimpleX (@ simplex.chat) v6.1 - "Vulnerability" to traffic correlation attacks, not surprising at all. Almost every protocol is, unless especially crafted in a way to mitigate that, which basically nobody ever does. Adding extra security for transport layer, hmm, ok. Feels like bit overkill, but sure, extra layers are extra layers. Using SHA3 instead of SHA2 is nice. Device compromise and contact information replacement attack, sure. If the end device is compromised then, uh it is. In general, still looking very good, compared to most of other messenger apps.

2024-04-27