XSF Discussion - 2017-03-05

arc: you should make a talk about that :-)

spam pointing me to a jid on antispam.im... oh the irony (-:

totally makes sense for spamproviders and antispam providers being in the same boat...the same works for cloundflare, who provide DDoS protection for DDoS booter sites ;)

*booster

Is there a reason muc doesn't force all resources into using the same nick?

I've honestly never considered that… I kind of want to go try it just to see how things behave.

"boom"

the reason i'm asking is because ejabberd now supports merging and parting multi sessions nicks. and Conversations behaves pretty badly when you actually try it

and gajim is even worse

which kinda lets me believe that this is a problem that can not really be fixed

you could fix it in conversations, couldn’t you?

Where "this" == "MUC"

also, where can I test how bad pidgin breaks?

I'd be curious to see how other clients behave though; theoretically clients have to accept whatever NIC the server gives them, so I'd *think* this would "just work", but never underestimate client devs ability to ignore the spec :)

Nick, even.

jonasw, yes and i'm about to. i'm mostly done

however that's like soo many edge cases to consider

i'm confident that *I* will get that to work somehow. but i'm not sure for like 99% of the other clients out there

in that regard forcing a clients into the same nick might actually be the smoother edge case

On a similar note, why don't servers let you change to an existing nick if it's also owned by another resource? Every time I try to change my nick in Conversations ot a new room I'm in from the default one it sets I get the dreaded "nick already in use", and then have to part/join to use my own nick.

It seems like this could work; but most of my intuitions about MUC end up being wrong, so…

if I understand you correctly, that’s one of the thigs ejabberd now allows and it’s simply because of "we don’t implement that" when it doesn’t work

SamWhited: yes I think that's a recent thing that ejabberd allow that now.

Prosody still prevents this

Excellent; all MUCs need to switch to ejabberd. That's quite possibly my biggest annoyance when using Conversations, I'd never considered that you could fix it on the server with either of these changes.

I think Zash mentioned that it’ll be a thing with prosody in 0.10 too

but I might be wrong

SamWhited: the nick already in use is a server error

Meaning you have to fix it on the server side

That's not a Conversations check

Yah, I'd always wanted to fix the symptom, not the problem (by having a "default nick" setting in Conversations)

I'd still like that, because I'd still like a default nick on servers that don't force all resources to the same nick, but it would definitely be better if the error were fixed on the server too.

Although, even if you force all clients to the same nick, nick changes would still be break things I guess because I don't *think* there's a way to force clients to change their nick if they don't request it.

SamWhited: xep 45, around example 51: If the service modifies the user's nickname in accordance with local service policies, it MUST include a MUC status code of 210 in the presence stanza sent to the user. An example follows (here the service changes the nickname to all lowercase).

and now that i fixed most of the bugs in conversations i think there is a bug in ejabberd

not clear whether that can happen out of the blue

Holger, when i'm joined with two clients A and B both using the nick x and client A changes the nick to y client B will see y join (which is correct) but client A wont see x join

which i think would be the correct behaviour...

but the fact that i'm not even sure about that doesn't speak for MUC

ugh

another thing you wouldn't have to worry about when forcing all clients in the same nick; if you you receive a message from another client with a different nick; does this count as a sent or as a received message?

ugh

can we have MIX please.

in Conversations it actually counts as a received message; which totally confused me when testing multi session nicks; but in general this is probably expected behaviour?

daniel: it's a received message, and: does forcing nick mean that I can no longer join a MUC with different nicks per resource?

kaboom, that whati mean by forcing the nick

but, why?

I use that as a feature...

kaboom, because it would get rid of *a lot* of problems

kaboom, https://xkcd.com/1172/

Is there any other big multi-user chat system that allows different clients using the same account to show up as different nicks? I don't know of any.

you can already have multiple resources share a nick optionally, why force it?

> kaboom, because it would get rid of *a lot* of problems

what problems?

those i just described

There are a lot of corner cases on the server wide with nick sharing

Prosody just doesn't let you change nick if you are sharing it with another resource, decided the complexity to implement it correctly wasn't worth it for the one person that might one day need this feature :)

MattJ: I'm sure. However I'm starting to assume that (forced) nick sharing has fewer corner cases than allowing client to merge and split their nicks

I'd potentially be ok with that (forcing you to use a nick you are already using), but I don't really see a need for it

It's probably what users want in (almost?) all cases

The main problem is that from what I've seen, clients haven't traditionally been very good at handling server-forced nick changes

MattJ: well I'm kinda starting to assume that this will be easier for clients and maybe even for servers. Gajim for example has completely broken behavior when joining and splitting nicks

I'm not sure why clients should have any issue with it, if the server does it correctly

prosody's current behavior makes it really hard to change your nickname if you are using two clients, already.

s/hard/impossible/? (unless you rejoin)

MattJ: yes, unless you rejoin with all clients at the same time. it's a PITA

Oh, you want to change nick on all clients

Yeah, we could implement it, but again, is it really useful? and then... will clients handle it?

It might be a good thing to enforce the nick change over all clients as well

I'm sure clients will love to receive a new nickname

Exactly

I still don't see how any part of MSN can negatively affect the client-side protocol implementation

No, I think it can always be transparent to clients if done correctly

But it's really hard to do correctly

https://share.gultsch.de/daniel/2QCLUVwufHdGb2xM/7yWOlGliQhCohcZwJlpvdQ.jpg

My gajim has a different story to tell on that

what MattJ said.

daniel: it's a received message if it comes from another nickname, and it's a sent message if it comes from another nickname. just ignore the non-anon-MUC JID ;)

The server has remote code execution on the client for the following operations: nick joined, nick changed, nick left

daniel: and what you described above sounds like a bug in ejabberd

i.e. it has full control over the nick list

in every client

the thing with "sent" messages in a MUC is: how do you know if it's the one you sent right now or a message from an MSN client?

To implement every operation for nick sharing, you have to stop thinking about room state management and code for client state management instead

MattJ: you need to think about both, actually.

Still, I'd love to have prosody either implement nick-split and nick-join, or to enforce a nick-change over all MSN resources

the current situation is really cumbersome, especially with clients that disallow editing a MUC you are not joined to *cough*conversations*cough*

I think enforcing nick change is likely the easiest, but I don't think many clients handle it

Maybe we should just try that and see

MattJ: +1 to that.

MattJ: I'm sure it affects the same clients that fail to recognize that your nickchange was rejected by the server.

MattJ, if a client doesn't handle nick changes; how can you expect them to handle the merging and seperation of nicks properly

daniel: because merge and separation are join and leave events for a different participant

daniel: when you initiate a nick change, you see that change followed by a join of your other client, everybody else (including the other client) will see your new nickname joining

daniel: I haven't seen any client handle forced nicknames

Zash, conversations certainly does. I haven't check others

sure that's not common?

daniel: all I tested broke in varying ways

Zash: now I'm surprised. They break on forced nick change, but not when the server denies a change?

Nick override on join at least

My client doesn't even always recognize MUC events.

Something is leaking presence listeners...

Just for the record: because most clients an servers fail to properly implement nick split/join, you want to remove the feature of being present with multiple nicks completely. Ever thought of removing just the split/join and forbid nick changes when being online with multiple resources? Solves the problem without losing a totally unrelated feature...

kaboom: prosody forbids nick changes for MSN, and it's a usability nightmare

Error> xsf@muc.xmpp.org/Zash: modify: Unknown error

Hm. something is wrong

Didn't I disable mod_firewall?

21:18:06 IN <message type="error" id="3ca0c0b6-3334-45f3-a65d-5d5d4a974adb-8D9B" to="georg@yax.im/9f1e6da6-2e3c-46d5-80ab-e73ac32d4719" from="xsf@muc.xmpp.org/Zash"><error type="modify"><policy-violation xmlns="urn:ietf:params:xml:ns:xmpp-stanzas" /></error></message>

I wonder why Zash isn't kicked for sending errors to the MUC

Only certain errors are "kickable"

The list isn't standardized

Yay.

Don't dig any deeper

Kicking on error isn't standardized, is it?

No, I don't think so

is there even one place in XEP 0045 that is properly standardized?