XMPP Service Operators - 2020-05-03

MattJ: Why fork prosody instead of working together to make prosody better?

Maranda: how exactly does jitsi use XMPP? Can an XMPP user contact a jitsi user? I've never even been able to use jitisi, any time i go to an instance it locks me out and says "It looks like your using a browser we don't support, Please try again using Chrome or Firefox"

I dismissed it as more nonfree crapware under the guise of an open source license

It reminds me of in the 90s websites

"This site best viewer with Internet Explorer 5

Or

We don't support any browser other than internet explorer

It's anti-competitive and flat out blocking people based on a useragent is retarded

I figure that it reflects the developer's mentality. That since they are blocking non-big-tech browsers that they don't care about freedom and interoperability and in fact are hostile to it

https://upload.nuegia.net/ca9b8918-2d96-4739-ac04-3fe950f06aae/screenshot.png

What kind of support are they talking about anyways? For something under an open source license there is usually a disclaimer of liability

I wish there were more native clients that supported jingle audio/video. Then I wouldn't have to deal with this anti-competitive web cancer

At least in this case it's not saying ONE MORE STEP! Please solve endless captchas but the end effect is still the same. Denial of service

Does that browser support modern specifications like WebRTC? I'm not sure what the cause is but I would first assume that that browser just doesn't have the required technologies built that that the web version of jitsi requires in order to run. Or am I misreading?

Yes

Also, you're trying to visit https://jitsi.org I presume?

I tried that

And no addons or settings or blocking anything important?

But currently i tried a privately run instance

I've never had any issues with the official jitsi program in the last several years.

Jitsi Meet uses XMPP but it's not a general-purpose client

Jitsi desktop does!

Though, Jitsi is trying to spread alot of awareness to it's newest Jitsi Meet program

I never really liked their desktop client, and the team don't really work on it these days anyway

tom, so, re Jitsi Meet and browser support: they rely heavily on WebRTC and some stuff. The choice to ban certain browsers is to ensure a good user experience for everyone else. There was a time (and that may still be the case unless you’re using Firefox 76), where any firefox version would cause massive problems for every other user

That's really a shit thing to do, regardless of the justification

No offense to you jonas’ and I appreciate the explanation

But if what your saying is true that just someone joining with a specific version of a specific browser is enough to crash the server, that doesn't speak well for the quality and reliability of the system

That's really reactionary instead of proactive

tom: the problem was a lack of support in Firefox for certain WebRTC features

The choice for Jitsi would be to make it only work in Chrome/Chromium or allow Firefox anyway and accept the instability (so they made it display a warning)

The next couple of Firefox releases are supposed to fix the remaining issues I believe

And some of the others the Jitsi team have already figured out ways to work around (such as bandwidth estimation)

tom, it’s not crashing the server. the effects range from "massively increased bandwitdh use due to lack of firefox features" to "audio breaks even for non-firefox users because of webrtc stuff"

tom, those were at least the issues with "older" (pre-76) firefox versions, and I can fully understand that you lock that type of software out.

and of course, you can always install your own jitsi meet deployment (been there did that) and remove the restrictions there -- it’s just an array of allowed versions. ✏

but not allowing crap implementations on your flagship platform makes sense to me.

and they also (try to) work closely with firefox to improve it. it’s not like they "silently" lock out firefox, they file bug reports and provide info to make it work

it’s just a process which takes time; firefox devs are hit by covid just like anyone else, and before covid the interest in such platforms was lower by orders of magnitude

tom: > MattJ: Why fork prosody instead of working together to make prosody better? That was done in 2009(?), it's not just a fork anymore

tom: > It's anti-competitive and flat out blocking people based on a useragent is retarded Video is heavy, tte browser needs to handle it

> tom, so, re Jitsi Meet and browser support: they rely heavily on WebRTC and some stuff. The choice to ban certain browsers is to ensure a good user experience for everyone else. There was a time (and that may still be the case unless you’re using Firefox 76), where any firefox version would cause massive problems for every other user Privacy or Video chat, choose: https://github.com/jitsi/jitsi-meet/issues/6230#issuecomment-616855160

tom: > how exactly does jitsi use XMPP? Can an XMPP user contact a jitsi user? Jitsi Desktop is a normal client Jitsi Meet...the users are in the MUC only for that video session not sure there's a need to be able to contact them, of course except if you want to use video

It's not a browser blacklist

It's a browser whitelist

They already support Chrome/ium (and I guess Vivaldi/Opera)/Firefox/Safari, you are using what?

tom, yes, because maintaining a blacklist with all those crappy chromium forks is not going to work ;)

Well it blocked me

because everyone and their dog thinks that they can write a browser just because they know how to slap a webview in a tab control

And I'm not using a chromium fork

Also

what *are* you using then?

That should be the users own right if they want to use a chromium fork

XUL

tom: erm Jitsi ain't Jitsi Meet btw

if it’s not going to work with jitsi meet and potentially causes annoyances/harm to other users, it’s their right to not allow you in.

Two different things

What that does though

It doesn't make people stop using it

tom: you have rights, the JM devs don't? Host your own, disable Simulcast, done

It just causes people to start tacking on a bunch of bs into their useragent string

Licaon_Kter: that issue is not even applicable to the XUL engine though

It doesn't even contain any Rust componets

tom: i don't follow, pls link to your browser

Just look at webserver logs of any chrome or safari browsed site

Mozzila 5.0, safari (not) chrome version blah blah (not really), internet explorer 6, firefox 666

Because web operators who didn't know any better started whitelisting useragent strings

Browser just started adding every variation of every browser useragent under the sun

Here

I will pull a log entry from my server

» [02/May/2020:REDACTED-0700] "GET / HTTP/1.1" 200 1379 "-" "Mozilla/5.0 (Windows NT 10.0; Win64; x64)AppleWebKit/537.36 (KHTML, like Gecko) Chrome/66.0.3359.181 Safari/537.36"

so?

Is it mozzilla 5?

Is it WIndows NT's internet explorer?

Is it KDE Konquer?

No

Is is firefox from the early 2000s?

Aka Geck

O

Is it Chrome?

Or is it Safari?

Nobody even knows anymore. It's just insanity.

Windows NT 😂

To give some reference for what a Useragent /SHOULD/ look like if supplied even at all:

[03/May/2020:REDACTED -0700] "GET / HTTP/1.1" 200 821 "https://redactedrefererer/" "NetSurf/3.9 (Linux)"

The new Microsoft Edge is a fork of Chromium basically for who doesn't know already

tom, are you a web developer?

All this insanity in that one little HTTP header to work around webmasters using the useragent to supply different html per useragent match

Today, web browser even ship with a HARDCODED LIST of per-domain useragent overrides to bypass this

*web browsers

jonas’: systems engineer

why does it matter to you then?

Bandend stuff, although i do front end too when needed

tom: we know UserAgent is shit and it will go away Did I miss when you linked to which actual browser you use?

Licaon_Kter, "XUL", so I suppose a crappy pre-rust firefox fork

Licaon_Kter: https://git.nuegia.net/webbrowser.git/

At least it's not Palemoon amirite? :)

tom: if you join (more than one, eg. One from Android plus one XUL browser) https://meet.jit.si/randomconferencename#config.disableSimulcast=true does not work? ✏

All HTML5 web browsers are terrible abominations. This one is just a little less so

Maybe instead of flat out blocking

The webapp could gradeful degrade and/or offer adjusting that as a suggestion?

it still degrades the service for everyone else

tom: feel free to PR, but iirc if one client does that it forces all members to do the same

Ninjaed :))

it doesn’t force everyone to do the same, but everyone will feel the negative effects

That...yes..

Licaon_Kter: I would if there's an actual address to send PRs and patches to

tom, https://github.com/jitsi/

Did you know that you can't even upload .patch files to github?

jonas’: nope

there’s the jitsi-meet repository with all the web frontend stuff, also jitsi-videobridge and jicofo repositoires

Better link to meet directly....so many apps there with SIP and all that :))

jonas’: some bloated webapp with 5 megs of google spyware javascript requiring you to sign up for a microsoft account to contribute to an open source project is not welcoming community contributed patches

of course not

tom: send patches by email as usual

Yes

Either a maintainer email box or a mailinglist or a bugtracker like bugs.debian.org

Future versions of Chrome are not even going to supply a useragent http header. Good riddance to. Exactly the kind of drastic thing we need to put a stop to useragent string discrimination

And everyone will apply the Chrome quirks? Yeah..fun...

https://developer.mozilla.org/en-US/docs/Web/HTTP/Browser_detection_using_the_user_agent

Even pozzilla in all their bad judgment recognize how using the useragent for different html is wrong

And no Licaon_Kter

The useragent header is not mandatory according to the HTTP spec

It's only purpose is marketing

Like X-Mailer header in email

Or not disclosing what client your using in XMPP

Not not responding to CTCP VERSION in IRC

It's an HTTP thing not an HTML thing

You can actually go all the way back to Fortran and how people got burned and learned to stop using proprietary vendor extensions to Fortran

Licaon_Kter: the plan is that hopefully ounce chrome takes the first plunge not supplying useragents it will give everybody else the ability to as well

And without a practical way discriminate browsers, people will use graceful feature degradation

Example being <script><noscript></noscript></script>

I'd love to see video conferencing implemented in <noscript> tags

That's probably not possible but i have seen interactive chat in noscript

Which is pretty cool

hmm aw mod_s2s_keepalive loaded on (muc.)xmpp.org as well :/