jdev - 2025-05-31

I have the most interesting idea for an animated group chat. gemini://woodpeckersnest.space/~schapps/journal/2024-07-15-xmpp-activity-and-mood-as-means-for-mingling.gmi

https://xmpp.pimux.de/file_share/0683b456-5032-7600-8c52-1ad8b2f1cb7d/2024-07-15-xmpp-activity-and-mood-as-means-for-mingling.gmi

It could be a great addition to lure guests (i.e. annonymous logins) into XMPP, especially with Candy Chat, Converse, Speeqe, and Movim.

Does anyone here has knowledge of PJS? > Summer of Code 2008 - XMPP WIKI > The PJS project aims to create a reliable, correct and highly scalable XMPP server framework in Python, not unlike what ejabberd is to Erlang. It was started as an undergrad project at the University of Toronto in February 2008. https://wiki.xmpp.org/web/Summer_of_Code_2008#PJS

Schimon: What application can open those GMI files?

Martin. These are text files. GMI files are called GenText, and are intended for the Gemini browser and protocol.

Fennec just downloads them. Maybe config your webserver to serve those as text and not as file?

Firefox also Downloads them

I do not own the pimux.de server. I will change the file extension to TXT.

https://xmpp.pimux.de/file_share/0683b4f9-686a-7bd8-a25b-d7befe7e7468/2024-07-15-xmpp-activity-and-mood-as-means-for-mingling.txt

Schimon: That works, thanks.

You are welcome. Thank you for spending time to read it.

> Does anyone here has knowledge of PJS? >> Summer of Code 2008 - XMPP WIKI >> The PJS project aims to create a reliable, correct and highly scalable XMPP server framework in Python, not unlike what ejabberd is to Erlang. It was started as an undergrad project at the University of Toronto in February 2008. > https://wiki.xmpp.org/web/Summer_of_Code_2008#PJS Schimon: the chances python last touched in 2008 works in 2025 are 0 ↺

Hell the chances python last touched in January works now is pretty small

moparisthebest. I suppose so, yet that subject is still interesting, and there are many interesting ideas that can be implemented today.

what's the idea ? "XMPP server in python" ?

Yes.

>> Does anyone here has knowledge of PJS? >> https://wiki.xmpp.org/web/Summer_of_Code_2008#PJS > Schimon: the chances python last touched in 2008 works in 2025 are 0 I think you spelled 100% wrong. Code doesn't just spontaneously stop working ↺

I mean that seems like the worst idea ever to be but /shrug lol

moparisthebest: Not sure why you say that, in my experience there is almost nothing that gets removed from the language

Why not? An XMPPP server implementation in Python could be used as a software to easily create proof-of-concepts of XEPs, at the very least.

>> Schimon: the chances python last touched in 2008 works in 2025 are 0 > I think you spelled 100% wrong. Code doesn't just spontaneously stop working Every new python interpreter version breaks existing code, you can't use old interpreters from 2008 for all the security vulns and they wouldn't compile on modern systems... That's what I mean by "won't work" ↺

> Every new python interpreter version breaks existing code, you can't use old interpreters from 2008 for all the security vulns and they wouldn't compile on modern systems... That's what I mean by "won't work" Yes. This is a problem. ✏ ↺

You said last January and that's clearly exaggerated

We upgraded the python version in Gajim mostly for better type hints. Without type annotations I'm sure Gajim would run fine with python 3.9 or something

And I don't think I've ever seen a new version seriously break existing code

Running new code on old python versions gets "broken", sure. But the other way around?

They remove obscure modules, so yes if you use these it's breaking, but like with 4 versions deprecation before

Constantly changing method signatures and such, poezio in Debian stable was just broken because of this the other day

https://www.sciencedirect.com/science/article/abs/pii/S0164121225000056 here have a paper lol

Anyway python is also way too slow to make a real server in, see synapse

And yet our best server is in Lua and our second best is in Erlang. Turns out servers aren't CPU bound so using a "slow language" isn't a bottleneck really

Lua and Erlang are both probably an order of magnitude faster than python

If I even bought into the idea that language could be fast or slow, that seems obviously untrue

hold up you don't think some languages are faster or slower than others? Seems absolutely absurd

> And yet our best server is in Lua and our second best is in Erlang. Turns out servers aren't CPU bound so using a "slow language" isn't a bottleneck really Synapse is CPU bound because it does so much computation, but an XMPP server wouldn't need to do so much ↺

> hold up you don't think some languages are faster or slower than others? Seems absolutely absurd No. Implementations can be but languages are abstractions with no mandatory performance requirements ↺

JavaScript used to be "slow" and now it's one of the "fastest" high level langs with no changes to the language itself in between

> Most software do not need “infinite scalability.” They need: > Predictability > Observability > Reasonable resource limits > A working dev/staging environment https://sliplane.io/blog/serverless-is-a-scam

> JavaScript used to be "slow" and now it's one of the "fastest" high level langs with no changes to the language itself in between It this change of speed of JS due to interpreters? ↺

Actually, the minimum memory consumption of Python is significantly lower than it was. KaikOut consumes less than 10MB, which is good.

https://xmpp.pimux.de/file_share/0683b579-6971-7478-b445-bd47d48c0296/slixfeed-scalability-2025-05-13_07.04.59.mkv

> https://xmpp.pimux.de/file_share/0683b579-6971-7478-b445-bd47d48c0296/slixfeed-scalability-2025-05-13_07.04.59.mkv This is Slixfeed, processing 30K entries. No multiprocessing neither threading. It only utilizing async and AsyncIO. ↺

From ralphm's journal. > Being at XTech again is great. My presentation wednesday might not be the best ever, and the room was way too large. If all people visiting XTech would be in the room, it would still have looked empty. On the other hand, speaking to people in between sessions and in the evenings was again inspiring and most people really like the idea of publish-subscribe. Also the suggestion of integrating Jabber facilities into a browser proved to be thought provoking. *Also, again, people are dying for a Twisted Python Jabber server implementation.* http://ralphm.net/blog/2006/05/19/xtech06

> Why not? > An XMPPP server implementation in Python could be used as a software to easily create proof-of-concepts of XEPs, at the very least. i dont see it being easier enough than lua to be worth writing another implementation ↺

part of why i run prosody is so that i have easy tinkering down the line

> It this change of speed of JS due to interpreters? yes, theyve done a ton of crazy shit both with interpretation itself and how good and quick jit is ↺

actually jit alone was probably one of the biggest things for js, and js's jit is one of the best

its still an interpreted language at the end of the day and far from "fast" but its been given possibly some of the most attention of all the languages particularly in terms of the runtime

Yeah. It turns out if you parse C source at runtime line by line an interpret the resultant ast directly it's pretty slow too ✏

very true

python does have some strange footguns when it comes to performance, but it doesnt really worry me. i more just question the utility over something like lua which is already a very high level interpreted langauge

however im a python novice and have never written lua so i would not know myself

Neither is my favourite but everyone has their own preferences and more is better

true

im all for another server implementation, i just think its worth thinking through

i actually kinda wanna see a server in rust or go, not because i need to run it (my prosody is far from bottle-necked) but just because