jdev - 2024-09-19

singpolyma, https://www.rfc-editor.org/rfc/rfc5122.html#section-2.2 > Certain characters are allowed in XMPP node identifiers and XMPP > resource identifiers but not in the relevant portion of an IRI or > URI. The characters are as follows: > > In node identifiers: [ \ ] ^ ` { | } > > In resource identifiers: " < > [ \ ] ^ ` { | } > > The node identifier characters are not allowed in userinfo by the > sub-delims rule and the resource identifier characters are not > allowed in segment by the pchar rule. These characters MUST be > percent-encoded when transforming an XMPP address into an XMPP IRI or > URI.

so "xmpp:foo\40example.com@otherexample.com" is not a valid IRI

because \ is not percent encoded

lovetox: sure that's possible. But the browser is percent encoding it anyway so that's not the main source of the issue here

Gajim does treat it like a JID, and not as a IRI I fixed that

> I thought you mean eg xmpp: someone%40example.com which is ok if course I mean, it probably works, but technically it's not right.

its not right i think, and Gajim would not parse this

@ as delimiter between node and domain, is not to be percent encoded, this would prevent a parser from splitting the jid into its parts

you can percent encode every single char in a URI if you want

a correct parser will decode every encoded thing

and so it will still be an @ by the time it hits the jid parser

dwd: In what way would you think it's not right?

i was talking about a IRI

thats what we use normally in xmpp environment, no need to convert something to a URI

Gajim has no support for the URI form

Those are equivalent for this purpose, IRI vs URI. They use the same escaping

no seems not if you say you can percent encode everything on a URI

the IRI spec is quite clear to me, and it nowhere says to percent encode the @ in an xmpp address

there are to capitals on processing and iri and generating an iri, and all it says is to encode by inode, ires, ihost rules

and respectively decode by these rules, but if you encode the @ you cannot get node and host, because jid splitting does not work anymore

I don't understand. The @ is not special at all. Why wouldn't you be able to encode it?

You split on the @ once you have the jid you are parsing. That's not part of the IRI syntax

what do you mean by its not special? there is a syntax to create a iri, and it says to place a @ between node and host

nowhere does it say to percent encode it later

Ah I see, rfc5122 makes it special even though it normally wouldn't be. Interesting. So if you specifically support xmpp IRI then technically it's special even though if you use a generic IRI parser it will work. Hence dwd's "probably works but technically wrong"

i implemented the iri syntax because thats what we usually use to share, URI make not much sense in context of sending xmpp links. And there is a special paragraph of transforming and IRI to URI and back, but i read that, and it was complicated enough that i didnt want to consider it

the problem with that is that most implementors do url_encode("xmpp:" + JID)

and think thats valid

IRI vs URI isn't relevant in most contexts yeah

What's different here is generic IRI syntax vs rfc5122 syntax

yeah i dont understand the use case where it would, so i would try to agree that we all use IRI, or someone eli5 to me how that conversion from URI to IRI works :)

URI vs IRI is just about Unicode or not

ah, because IRI allows unicode chars

and URI percent encodes them all

Yes

That's all