XSF Discussion - 2019-04-10

flow, I suggest message

you could make use of errors with error type='continue' along the way to introduce the traceroute-like backscatter...

jonas’, yep, that was one reason to use IQ in the first place, but as IQ has no advantage over message with the correct type, I now also lean towards using message

although I don’t see too much harm with modifying an IQ in-flight in a well-specified manner. however, that requires some type of support in the recipient when scattering it back.

implementing the backscatter with error stanzas sent by the hops themselves might be nicer

hm

IQ has the advantage that the recipient will send you back at least service-unavailable

so you know when to stop to listen for backscatter

which makes me tend slightly towards IQ

I'm not sure what the use case is, but having a generic stanza payload element that tells the server to send back a dedicated message with the annotation sounds like a more reasonable approach to me

What if you never receive the actual pong?

All trace route info is lost tung

Then

Ge0rG: What if you mostly receive the response, either in form of a pong or of an error type stanza?

jonas’, isn't there a message type where an intermediate hop will send an error back, ideally with the full payload, in case the stanza can't be routed?

flow, not built-in

flow: what if you want to debug _why_ you don't receive a given response?

jonas’, groupchat maybe?

flow, no

Ge0rG, then I could still go the traceroute approach and proceed with the next closest hop in case of no response

flow: oh wait. you can just send three ping IQs to all the hops!

flow, how would you know the next-closest hop?

it's xmpp! it's got a fixed routing structure.

my full jid -> my domain jid -> their domain jid -> their bare jid -> their full jid

jonas’, ahh right there is a case where groupchat does not trigger an error respone: RFC 6120 § 8.5.3.2.1.

Ge0rG, metre? :)

Ge0rG, components?

jonas’: how do you obtain traceroute results from a thing without a JID?

it could have a JID, but not the one you expect

In theory, I could assign my metre-like thing a JID for management/monitoring purposes, and that could be the JID which replies

jonas’, both approaches have their advantage and disadvantage. I would lean towards the initial one, but that doesn't mean that we could also do the other one, i.e., by having an extra flag which indicates that intermediate hops send their data right away

is there any reason for intermediate hops to *not* send their data right away?

well if everything works according to the standard then they don't need do

that’s wrong

a recipient is not required to mirror the payload in an error message if it doesn’t understand it

it is a MAY or a SHOULD, but not a MUST

so if your final recipient does not understand the IQ based protocol, you won’t see the intermediate hops

which is meh

> well if everything works according to the standard then they don't need do well if everything works according to the standard then you don't need traceroute

Ge0rG, I don't think I would conclude that

jonas’, that is also true mostly for the message based protocol

flow, which is why I’m leaning towards IQ and backscatter with type='error' type='continue'

jonas’, yeah that is a strong argument for IQ

"matrix.org is not recognised by the IETF as an Open Standards body" - https://mailarchive.ietf.org/arch/msg/art/EjDTW6KhrOYebVS8SGsZh7BEE9g

I am probably going to switch xep-traceroute from <ping/> to a custom IQ

FWIW, I have no idea what Metre would do in terms of a traceroute thing. It's intentionally invisible, beyond testing cases.

flow: your proposal lacks a rationale so I can't say what you need it for when everything works

I don't think I proposed something. But besides that, there is a short and possible incomplete note in the introduction.

So what is it needed for when everything works?

You probably shouldn't use something like traceroute to test if a channel to a remote entity is encrypted. But it sure would be nice, and if it is just to satisfy my and possible other people's curiosity about how the channel may look

Is it possible to have asymmetric paths in XMPP?

jonas’: btw, responding from an "unexpected" JID most probably won't work because you need to s2s auth that domain first

Ge0rG, With Metre and stuff like it, anything's possible.

Ge0rG, what stops such a thing from actually having that JID available via s2s?

e.g. gateway.company.example, could easily have s2s records :)

jonas’: most probably: nobody testing for it

jonas’, One thing that stops it is that nobody would want to expose such infrastrcture without good reason.

That, too. Nobody wants to expose themselves to xml deserialization.

Oh. Wait.

completely off-topic, but this is the channel with most French nationals that I know of - is a company named AYOMI legit? Seems to be a crowd-funding platform, but Gmail warns about phishing.

dwd: What is an "Open Standards body"? Those exact words are not in the RFC referenced? ✏

Zash, No idea. It just amused me. The XSF *is* considered an Open Standards Body, as far as I can tell, because otherwise we wouldn't have got "urn:xmpp:"

https://tools.ietf.org/html/rfc2026#section-7 I assume

someoone receive my photo

https://bgzashtita.es/upload/2648c95f187eff879a53c9890338a1a8d74210e6/5K0I4twyQ0gldJzYDaGx1guJq9vj23dsGsRMeIJm/20190410_145407.jpg

Is this relevant to XMPP standards?

I see the picture.

Guus, thank... disroot didn;t see it.

is is possible to bridge SIP and XMPP ?

Probably

Also see https://tools.ietf.org/html/rfc7573

edhelas, Do you mean bridging SIP and Jingle, or SIMPLE and XMPP-IM?

both actually

maybe the SIMPLE/XMPP-IM could be a nice first step

edhelas, So i think all the standards exist, but implementations, on the other hand, are harder. Asterisk might help for voice etc, I'm not so sure about IM.

edhelas, I have a very vague recollection that Openfire has a plugin, too - no idea what shape it's in.

ok

There is a SIP plugin for openfire that is still very much in use. It maps the SIP credentials to an XMPP JID. I use for users and even SIP proxies

nice

Would this make sense? (poke jonas’) $(all_xep_xmls): $(OUTDIR)/%.xml: %.xml - cp $< $@ + xmllint --nonet --noent --loaddtd "$<" > "$@"

Eg with how https://xmpp.org/extensions/xep-0226.xml can't be parsed by tiself because external entities

Zash, make a PR, I’m happy to test it

Zash, hm, instead you could copy xep.ent into the build/ too

MAYBE

So much indirection in the Makefile ✏

Metafile

*hick*

Somebody killed the bridge.

burned?

Now if I can never leave a bridged Matrix room, can they leave bridged XMPP rooms?

I recall there being a question/issue about that in the tracker

https://github.com/matrix-org/matrix-bifrost/issues/63 maybe?

Hm, wait