jdev - 2022-02-13

Hello, I'm trying to implement some xhtml-im in Renga so I can get irc colors and styling from Biboumi. But I don't see the xhtml in the messages I receive. I am testing with messages stored in the backlog of a muc (no mam implementation yet in renga), could that be a problem? What else should I check?

I have added the namespace to my disco info (I think), and it doesn't seem to help

I didn't know biboumi did xhtml-im

https://github.com/louiz/biboumi/blob/master/src/bridge/colors.cpp A very small subset of it, to render some of IRC formatting (foreground and colors, bold, italics, underline)

pulkomandy: last I checked biboumi did appear to strip all XHTML-IM from reflected messages

Maranda, it doesn’t parse XHTML-IM from XMPP clients, but it does convert IRC formatting to XHTML-IM.

pulkomandy, I have never tested it from the MAM history, but looking at the sqlite schema it doesn’t seem to store it.

Link Mauve: well but if a xmpp client sends a message containing XHTML-IM is it supposed to reflect it back mangling the stanza and stripping the XHTML-IM content?

Maranda, yes, since it couldn’t parse the XHTML-IM it just reflects the body.

So that you don’t think it worked and have the same view as other IRC users.

Does it preserve the id at least?

Not sure.

IRC also doesn’t have this concept.

Which is not exactly a bug... But biboumi not preserving id on reflection and mangling the stanza completely confuses deduplication indeedly ✏

Maranda, an issue with that is how to handle split messages?

Say you sent a message over two lines, and got a reflection of the first line in one message, and the second line in another message.

What should you do? What should biboumi do?

Huhu but is there actually a need to split out reflected messages on the XMPP Horizon...?

Well, this is about IRC, not XMPP.

IRC doesn’t support messages longer than 512 bytes (including your recipient name, some command, and the phase of the moon).

> pulkomandy, I have never tested it from the MAM history, but looking at the sqlite schema it doesn’t seem to store it. That's a bit unexpected, so MAM loses all formatting and possibly anything in a message that it doesn't support? Not really what I expect from an archive?

Actually if the IRC colors are stored in the body in the table it might, but I have no easy way to test MAM here, especially not with formatting, since I have no other IRC client than biboumi.

Link Mauve: well if I send a message longer than 512 bytes it's fine if it gets split on IRC, less so if the reflected message by biboumi MUC interface gives me back 2 messages instead of one, not that it's forbidden or anything by the spec that I can recall.. but for sure that may confuse some clients.

Maranda, you’d be confused as well if you got kicked for having flooded a room, and not understand why because to you it was a single message.

Well if biboumi does mangle the stanza, doesn't preserve the id, may not preserve the message body contents and even split out reflections, I got nothing to match on 😆

Oookay

ircv3 has message ids link mauve

Might be useful, could you mention that on the biboumi bug tracker?

Do IRC servers use this extension?

I don't know, i have not investigated it much

https://ircv3.net/specs/extensions/message-ids

Where is the bugtracker?

At https://lab.louiz.org/louiz/biboumi/-/issues

You need message tags enabled to use that spec anyhow, you could also use message tags to store xmpp specific data potentially

How does that work?

Is it a container for arbitrary payloads?

kinda, yeah

"Message tags are a mechanism for adding additional metadata on a per-message basis. This is achieved via an extension to the protocol message format, enabled via capability negotiation."

spec: https://ircv3.net/specs/extensions/message-tags

link mauve: any specific ticket? or just put the link somewhere?

I’m not sure, and I’m going to take a nap, yesterday has been tiring.

Sorry, if you haven’t found anything I’ll look for one in about one hour or so.

okay, I didn't know where you wanted that info, I'd have to dig out my gitlan credentials... so if you want to add it that is cool too :)

I confirm that I get xhtml-im for new irc messages but not from muc backlog. That makes my testing a bit more difficult, being able to re-check the same message would be nice

Now let's write a css parser…

I.e. attributes like `color="blue"` instead of `style="color:blue"`

I should learn how xml works, and see if i can make my formatted messages work with xml instead of json

so... the xhtml-im content is <html xmlns='http://jabber.org/protocol/xhtml-im'><body xmlns='http://www.w3.org/1999/xhtml'>[<span style='color:blue;'>HaikuArchives/Renga</span>] <span style='color:brown;'>pulkomandy</span> pushed <span style='color:green;'>1</span> commit to <span style='color:green;'>master</span> [+0/-0/±1] <span style='color:lightmagenta;'>https://github.com/HaikuArchives/Renga/compare/8d709ab45869...98879d2afdec</span></body></html>

trying to iterate over this with gloox I get one cdata string "[] pushed commit to [+0/-0/±1]" and 5 spans

not sure I can reconstruct the correct message from that :/

(that's one single cdata lumping together whatever is outside the spans)

Gloox 😰

it was working reasonably well until now

do I have to rewrite my whole client yet again using something else?

No, but I think just the mention spawns at least some decade old painful memories 😅

well I opened a bug on gloox bugtracker about it as I don't see a way I can reconstruct the message correctly from their API. And I'd rather not add another XML parser to my client (getting one to work right is enough of a problem, I think?)

In poezio, I chose to serialise the XHTML-IM body to string, and re-parse it with a SAX parser, which gives events in-order based on the position in the stream.

Very convenient for appending characters to a buffer as we go.

yes, I can do that, I guess. So are there any recommendations for an XML parser easy to use from C++ for this?

Most likely gloox already uses one, so you may want to reuse the same.

I think gloox does its own parsing, see Tag::parse in https://camaya.net/api/gloox-trunk/tag_8cpp_source.html

or is that parsing xpath quesries?

Looks like XPath indeed.

ldd doesn’t list any known SAX library though.

yes, xml parser is here: https://camaya.net/api/gloox-trunk/parser_8cpp_source.html

So maybe it does do its own thing.

Right. ✏

so I don't get anything lower level than the Tag class from that

Wut, why does gloox handle XML comments?

Lines 308-322.

pulkomandy, Tag::addCData seems to add it to both the m_cdata and the m_nodes members, you want to iterate on the latter.

Nodes in XML are either element nodes or text nodes.

but m_nodes is private and has no getter or iterator in Tag

Seems to be the issue you want to fix then.

yes, seems reasonable to add a way to access this if it's already there

ok, that worked :). Now let's add the css parser

Sam, re: host-meta tests, minimal nginx.conf's required: https://github.com/moparisthebest/xmpp-proxy/blob/master/integration/23-s2s-websocket-host-meta/nginx1.conf & https://github.com/moparisthebest/xmpp-proxy/blob/master/integration/24-s2s-websocket-host-meta-json/nginx1.conf pretty simple I think