XSF Discussion - 2018-02-13

“08:13:27 zinid> the problem is that now I'm not sure if other clients don't break though”, make it opt-in, for example with a disco#info feature, that way the MUC service will only send its own presence to clients which support it.

huh?

It was about MUCs sending a presence to participants.

Sorry, I’m backlogging.

This should be specified too, at some point. ^^

"opt-in"?

where guys you find these fancy words 🙂

none of my dictionaries know about it

optional I guess

zinid, in the MUC join presence, if the caps include this one feature then you can send the MUC presence, otherwise you have to assume it’s an older client and it won’t support that.

opt-in means "not included by default, unless specifically asked to be included"

zinid: btw. I've commited new muc roster in Psi. so it should handle it. But of course it may fail some other way with this new presences =)

zinid, https://en.wiktionary.org/wiki/opt-in

rion, ok

Same for streaming stream-features, if it were negociated it would be fine to send them as a “nonza” whenever they change.

my turn, what's a nonza?

SaltyBones: not a stanza

level-1 stream element like a stanza, but not a stanza

so any direct child of <stream:stream> that isn't iq/presence/message

SaltyBones, Nonza is a confusing made-up word that some people mistakenly think makes our specifications clearer.

:)

nonzas are rare elements that you can find in some streams, they are difficult to observe and some developpers are still discussing about their existence

SaltyBones, cf. https://xmpp.org/extensions/xep-0360.html

Dave Cridland, I don’t find non-stanza stream-level elements very useful either ;-)

jonasw, Maybe, but it didn't take an entire spec to explain.

Not sure if this is useful but for those who speak german or want to use google translate: https://www.kuketz-blog.de/messenger-matrix-das-xmpp-fuer-hobby-admins/

The article argues why matrix is better than xmpp.

Dave Cridland, I don’t think that we necessarily need a spec, but I think that having a term for that is useful; and in the end, all terms are kind of made up (like stanza); I’m not sure where your argument against that term comes from.

SaltyBones, Anything useful there?

Dave Cridland, the most useful thing, imho, is the statement that $all_his_other_problems are due to his incompetence at setting up prosody correctly.

Which, I agree, is an issue. Setting up xmpp servers is too hard.

I'm guessing bad clients are also an argument in that article.

is it though?

jonasw, yes? :)

SaltyBones, It is, in part, the SRV records.

you don’t even need those

Seve, of course they are, especially iOS

jonasw: yes it is.

jonasw, just remember jabber without mam and stream management...

SaltyBones, hm, good ponit

I always forget about the more advanced features

and setting up ejabberd...I mean I have it running but whenever I do more than use ejabberdctl I get stuck in the erlang shell :p

jonasw, the problem is they are not more advanced :) They are basic to the point of message loss without them

SaltyBones, yeah, I’m not saying that they shouldn’t be basic

but I always forget about them when thinking about how difficult it is to set up a server

(even though with 0.10, MAM should not be that much of a trouble anymore. isn’t it in core even now?)

Want to show you how to setup a server in under a minute? 0:-)

(sans SRV records, granted)

Well, yes? :)

Make a youtube video!

I'll post it under that article. :)

I actually made a video

15 minutes

awesome!

but it also gives you an EC2 setup and integration with our jitsi meet plugin

searching...

bah, we migrated forums, and that thread is lost to me

(more searching)

Oh

https://discourse.igniterealtime.org/t/installing-openfire-4-2-0-alpha-and-ofmeet-0-9-1/62240

when a video opens with VLC, which you virtually never use and you’re confronted with this: https://sotecware.net/images/dont-puush-me/EjZ7corpc2V-3hCD-slkA9f4XxK0w4uxoWP1BapKB3k.png

"Don’t blindly follow these instructions, without giving thought to security, interoperability and performance."

Guus, how about starting the video with a quick demo of what the end product will be able to do?

SaltyBones: obligatory disclaimer.

jonasw: yeah, it can be improved upon. But this doubles as an instruction for one of my customers that needed the EC2 bit as well.

jonasw: also: I like that ascii-art encoding on vlc :)

Guus, I can post this but I still think a 1 Minute thing would be very cool. :)

SaltyBones: yeah, I think this one is to long

I mean...not sure if you can actually do it in a minute .. ;)

I'm 5 minutes in, and I am explaining about nightly builds :/

SaltyBones, hm, I have ansible roles which work on debian jessie and stretch for prosody (both from package and from hg)

maybe that’s worth sharing?

but I think I write roles in a style which is very different from what everybody else using ansible does, so I’ve always been hesitant to share them

SaltyBones: 1 minute is hard while talking about it. But, to install: download platform specific installer, double-click installer, open browser and visit http://localhost:9090, fill out 4 screens, done.

depending on your download speed, you can do that in 1 minute.

Guus, remind me tonight and I will try ;)

in my video, skip to 7:30

jonasw, if they work I think it would still be nice

they *do* work. for me. they depend on my own homebrew ferm role though

> Which, I agree, is an issue. Setting up xmpp servers is too hard 1. Postulate that servers are complex in XMPP by design 2. Cry they are hard to configure

zinid, browsers are complex too, but need 0 configuration.

your argument is invalid :)

jonasw, yeah, that's why I have to always configure them

jonasw: my browser needs a lot of configuration to be barely usable

daniel, sure, but you’re a poweruser

also, you can install a server by entering a hostname and never configure it, no problem

most what other people do is install an ad-blocker if anything and be done with it

half will not be working, but the same is in browsers

Wow

I wonder how you use browsers that you have to configure them

Seve, and I wonder how the comparison holds?

you need DNS, opened ports and stuff like that for xmpp server

you don't need all this for browsers

ah, and certfiles

Indeed

also, when I often hear "hard to configure", that means the software is compiled from master branch, installed in some obscure way with an obscure configuration file from stackoverflow

But ejabberd is fairly easy to install these days

well, it still doesn't have gui configurator, so a lot of admins pass by

zinid, to be fair, when I see a gui configurator, I pass by

:-)

I was about to say the same thing

jonasw, that's why you need to define audience first 🙂

I don’t know how matrix is configured

click on the link ;)

None of this is going to solve the other problem shown by Roland. UI inconsistency and horrible iOS user experience

SaltyBones, that’s not good for my blood pressure

I think the recommended approach for the matrix server is to curl pipe something to sh to install node and npm

And then do some weird node magic to install something that might still be missing dependencies

And that will break on the next OS Upgrade

daniel: that's true for 90% of all node.js projects, for me at least.

daniel, it seems he is using python stuff

don't see any node in there

daniel, so we need to provide workable docker images

https://github.com/matrix-org/synapse <- python, not js

So who is talking to the KDE people to convince them of XMPP?

Ge0rG, why to convince them? You cannot convince IRC users to use XMPP

Ge0rG, Seve is

zinid: please shut up.

Ge0rG, fuck off

are you two going to have a fuck off-off now?

jonasw: can we have some thimbleweed and Ennio Morricone please?

Ge0rG, I sent an email to JUser mailing list to see if I could gather some interest and help on writing some document showing what XMPP can do and what do we have in terms of clients and such. I would like to tell them that using XMPP is also safe bet on the long run and such things. Unfortunately, I've been taking a look at these fancy clients like Rock chat or stuff like that, and we are a bit behind...

"a bit"

Yeah...

Seve: but we are not run by a single company, and the KDE community is full of developers. We just need to market it as a challenge for them!

Seve, I'm wondering if the JUser mailing list is being used at all, these days.

the log files seem pretty empty: https://mail.jabber.org/pipermail/juser/

Ge0rG, full of developers who have their plate full with their own challenges already, I’m afraid

Seve: I fear writing to an abandoned list won't move us forward.

Seve, I suggest you (also?) mail to the members list, or jdev, depending on the subject.

Guus, I think operators and members would be most appropriate.

maybe developers, too, but operators have the practical experience

just send in all lists

Ge0rG, there's Kaidan https://github.com/KaidanIM/Kaidan Which I wanted to point out this, when writing to them, to let their community know about this project (which I could test at FOSDEM in a mobile phone).

This is not going to make a difference at all, but always worth mentioning them

Guus, Ge0rG: Yes, I knew JUser is not really used, but I was afraid of sending it to another list and listen to people say "WTF r u doin here"

Reading their requirements I wouldn't recommend xmpp to them

Seve, don't worry to much about that :)

daniel, why not? They list that the list does not distinguish between must-haves and nice-to-haves.

we're ticking an awful lot of boxes - and sure, we don't tick all of them.

no-one will.

> Reading their requirements I wouldn't recommend xmpp to them If "not locked by a single company" can be ommited, then there are plenty of solutions

I don't think we should persuade them to use XMPP.

Or, that shouldn't be our goal.

they already have xmpp-server at kde.org

Helping them understand whether XMPP can work for them for this, great, but using XMPP where it isn't appropriate shouldn't be a goal.

And helping us understand where XMPP isn't appropriate provides value to us.

Guus: the thing is I don't think they are looking for a protocol but for existing software. I wouldn't know what desktop clients to recommend

I haven't looked at their requirements, but I assume there's some stuff in there that Swift doesn't do, else that's an obvious Qt-based option.

daniel, that, I understand. Are they currently defining a KDE-spec IRC desktop client?

Kev, how many people are working on Swift (if not a secret)? Why is it being developed so slowly? 🙂

Kev [14:46]: > And helping us understand where XMPP isn't appropriate provides value to us. Do we have the personnel to collect that information and to act upon?

"act upon" by producing even more discussions!

my name is zinid, I hate XMPP and the XSF in particular, so I will hang out in their channel all day and make useless and aggressive remarks

Kev, looking at their list of requirements, an XMPP-based solution looks at least about as fitting as another solution. I'd definately not _not_ engage KDE.

SaltyBones, and what did you produce here?

it's what he doesn't produce that doesn't make him appear as counterproductive.

(did I use the proper amount of double negatives there?)

I don’t know

I've brought this up before: I do like us to have arguments, but lets try and keep this a non-hostile environment to discuss things.

vanitasvitae, there?

I am

"Also almost all examples in the SIMS XEP are lacking a 'uri' attribute, which is mandatory according to §3.1 in references." - I see lots of uri attributes in https://xmpp.org/extensions/xep-0385.html

what am I missing?

The outer reference is missing the uri attribute

Take for example Example 1

The references in the source element are having uris, but the outer references dont

vanitasvitae, missing the SIMS URI?

?

My point is, that the References XEP makes an uri attribute mandatory

ahh

i see what you mean

:)

:)

wonder what would make sense there. I think https://xmpp.org/extensions/xep-0372.html should have an optional URI attribute if there's further content inside the reference

don't know what Kev, author of References XEP, thinks about that

zinid: It's not a static number, because people work across all our XMPP stuff.

(I like how you referenced the author of the references xep there)

(Without using a reference)

my client wouldn't recognize that either way, so I wasn't sure :)

Tobias: You mean an additional uri attribute over the one that's already there?

Kev, https://xmpp.org/extensions/xep-0385.html#example-1 I reference a shared file there, which is further described inside the references tag. the outer references tag doesn't have a URI element

Ah, I see. Yes, if the 'referenced thing' is part of the reference, rather than external, not using a uri makes sense to me.

someone should update the references XEP accordingly

vanitasvitae, if you're planning to implement it, I think Movim supports it too :)

Kev, I see

What about origin ids in references?

I tried to clean up my notes about yesterday's discussion and posted them to the mailing list.

SaltyBones: great!

Memberbot is online for all XSF members

Wow, significant loss in reapplications

Ge0rG: ya

Alex: memberbot is not asking me if I want to cast my vote now

lovetox: we lost you :(

jonasw, separating the hashes in ecaps2 would only be beneficial if the features/forms of an entity change (frequently), or are the other cases?

Alex, daniel: it appears to be panicing

> Error message from server: SleekXMPP got into trouble.

I'm getting an undefined-condition from it right before that.

redis was not running on my server

sorry

please try again now

Always forget that the bot depends on Redis ;-)

Oh. And it looks my resource. I can't start voting on my phone and continue on my desktop. Very 00s 😀

I see only myself is this room right now

are there problems in xmpp.org, or is my client running crazy?

Ye ol' MUC desync?

Alex: you've attempted to rely on TCP. Sorry.

I found a missing redirection: http://xmpp.org/about-xmpp/faq/