jdev - 2020-01-15

does it make sense to try PLAIN on a host from xmpp-client?

or does xmpp-client always mean START TLS

PLAIN? The SASL mechanism?

no, i mean no TLS

Or do you mean a totally unencrypted connection?

unencrypted

It doesn't make sense to ever try an unencrypted connection

If what SRV resolves to doesn't offer TLS, it's a matter of local policy whether you'll continue without it.

try is the wrong word

the user wants to connect PLAIN, do i resolve xmpp-client and try there

Yes

or do i just resolve domain

and use 5222

You would still obey SRV

You always use SRV unless the user explicitly asks you not to (if you support such a thing).

i wondered because xmpps-client is explicitly for DIRECT TLS

so even if the user wants PLAIN, i would never try there

i wondererd if xmpp-client is explicitly for START TLS

Calling it PLAIN is somewhat misleading. PLAIN is a SASL mechanism.

You mean 'The user asks to never use TLS', I think?

lovetox, no, xmpp-client is for the TCP binding of XMPP defined bin RFC6120

yes, the user asks to connect unencrypted

ok good, thats what i wanted to know

so i still have to resolve xmpp-client, before i go the domain:5222 fallback route

Yes

lovetox, that is what is written down in rfc6120 ;)

i wondered why did we add websocket and bosh into a TXT record

why not xmpp-client-websocket._tcp ..

Because they need URLs

ah its not allwoed

ok

its bit weird because it doesnt allow for weight and priority

but i guess that would be really an edge case to have more than one websocket connection point

HTTP doesn't

ah so it always has to be over 80/443

Yes

that makes sense then

We were discussing recently the 20 year-old bug report for Firefox to support SRV for HTTP

Web folk have no interest in it

writing since quite a time to move all connection/dns logic from gajim into nbxmpp

if you want to support, xmpps, xmpp, websocket, unencrypted, proxy for websocket and tcp, setting a custom host (no srv resolution)

and the lib should handle that all automatically, does get a bit complicated :D

Indeed

Someone needs to make a flowchart

lovetox, that's exactly what I am also currently dealing with in my not-so-small smack refactor project

but i have the luck of GLib taking most of the dns resolution away from me

because if i would have to deal with stuff like happy eyeballs etc

then i think it gets a bit crazy :)

I think you probably want happy eyeballs not only on the DNS level, but also on the connection level

actually they ever provide methods like connect_to_service() where i just pass "xmpps-client"

e.g. initiate TCP connections to the ipv4 and ipv6 endpoints in parallel

yes flow GLib does that

i incorrectly said DNS resolution, but GLib also does my connection

nice

though i had to implement socks5 proxy myself, because the code is really broken in GLib, but socks5 is really minimal, so not a problem

now you may only want to make parallel connect_to_service calls for tcp+tls/direct-tls/websocket/bosh and use the transport whoever finishes first ;)

hm no because they are not equivalent

(I actually thought about doing that in Smack but decided against it for the moment, only the endpoint discovery runs concurrently)

lovetox, I'd argue they are equivalent

not direct tls has fewer roundtrips

websocket is not as efficient as tcp

sure, but why does it matter

why does efficiency matter?

that's not what I said

of course i will try the way with fewest roundtrips first

yes, there is a higher overhead in some transports, but isn't it also important to the user that the connection is established fast?

also the chance that xmpps connect is failing although the server admin specified a srv record for it

is probably minimal to 0

eyeballs for ipv4, and ipv6 was made, because some networks didnt allow for ipv6, and some did, so there was no way to tell at start what will succeed

so they race both

lovetox, why try the one with the fewest roundstrips first when you could try all concurrently and simple use the one that finishes connecting first?

because its much more work

and almost no benefit

It also possibly depends what you mean by 'connecting'.

opening a writeable socket

lovetox, right, that's why I decided against it (for now)

A server operator probably isn't going to thank you for doing 6* as many TLS negotiations as you need, etc.

Kev, true, but you could just not do TLS and define "connected" as layer 4 connection (e.g. TCP) established

yes only open a socket

also i do not retry with different connection method, if something fails while TLS negotiation or even on xmpp level

then its just failed, im not helping the admin shadowing broken connection points

well retry a different XMPP transport after a TLS negoation failure appears sensible. After all it could be a misconfigured BOSH endpoint while the other endpoint is correctly configured

yeah exactly, no it does not make sense to me

Or not misconfigured, but a local transparent proxy is breaking HTTP.

this way the point will never get fixed

lovetox, I usually follow that argumentation, but it's the job of a client library to establish a connection whenever possible

aborting the connection attempt seems even for me a little bit to harsh

i think we have to disagree here

of course we have

and i think you never implement this right?

define "this" pls

retry different connection method if tls or xmpplevel fails

I was explicitly only talking about TLS negotiation

No, since Smack past and current design does iterate over different transports. It's only the new design that allows for modular transports to be used for the same connection, so I was not able to implement it in the past

yeah retry on error on xmpp level is just wrong for me

and TLS .. maybe

but think its tricky, would have to find out what TLS errors are errors where i want to try different

cert error for example definitly not

I'd thought especially cert errors

and really do i want to connect desperatly to a server , where the admin didnt set up TLS correctly

i dont know

*does *not* iterate

yeah i need to implement also POSH

i forgot about that one

Hello all, I'm trying to implement a "reply to" requirement. I use an ejabberd server. Can anyone point me directions on how I can implement this? I know about Threads (xep-0201), but the problem is that I need to get the parent message from the server and I didn't find a way to query for specific messages (I don't have all the message history, because I'm paging it)

.

,

You mean that you have a message, it says it is "in reply to message XYZ" and you need to find message XYZ?

yes

I also could make a query starting from 'message XYZ' and returning all other messages after it

but the xmpp specs only gives me three fields: with, start, end

Yeah, the current version of the XEP doesn't really support this use-case

hack: ask for one message and set rsm 'before' to the one you want and 'max=1', then request the message after the one you get.

Ha

But no, that's a terrible solution :)

Zash , I tried that solution, but I would have to send a start/end filter. But the stanza only gives me an id which I believe is a unix timestamp. Still, I could not convert it to a proper datetime string accepted by the form_type. It gives me a NAN error 🤔

You shouldn't assume anything about archive ids, they're opaque strings.

As for ejabberd requiring all three MAM form fields, that's just weird.

it does not

Looks like it to me

maybe we talk about different things

but its impossible, the prime uscase and example in MAM is querying the archive without any formfield set

and im pretty sure i never did set all 3 formfields ever

Every time I try that with an ejabberd, I get an error back.

and MAM works fine

MUC-MAM tho

why would that be different

Don't ask me

in ejabberd (my settings) you could pass zero or more of the following fields: <field var="with" type="jid-single" label="User JID" /> <field var="start" type="text-single" label="Search from the date" /> <field var="end" type="text-single" label="Search until the date" /> <field var="withtext" type="text-single" label="Search the text" />

please show the xml of your query

sure, one sec

<iq id='2' type='set' to='kuketzblog@rooms.dismail.de'><query queryid='f850e845-ce9e-4fd1-875d-206c6394ddbe' xmlns='urn:xmpp:mam:2'><x type='submit' xmlns='jabber:x:data'><field type='hidden' var='FORM_TYPE'><value>urn:xmpp:mam:2</value></field><field type='jid-single' var='with'/><field type='text-single' var='start'/><field type='text-single' var='end'/></x><set xmlns='http://jabber.org/protocol/rsm'/></query></iq> <iq type='error' from='kuketzblog@rooms.dismail.de' id='2' xml:lang='sv'><query queryid='f850e845-ce9e-4fd1-875d-206c6394ddbe' xmlns='urn:xmpp:mam:2'><set xmlns='http://jabber.org/protocol/rsm'/><x xmlns='jabber:x:data' type='submit'><field type='hidden' var='FORM_TYPE'><value>urn:xmpp:mam:2</value></field><field type='jid-single' var='with'/><field type='text-single' var='start'/><field type='text-single' var='end'/></x></query><error type='modify' code='400'><bad-request xmlns='urn:ietf:params:xml:ns:xmpp-stanzas'/><text xmlns='urn:ietf:params:xml:ns:xmpp-stanzas' xml:lang='en'>Bad value of field &apos;with&apos; of type &apos;urn:xmpp:mam:1&apos;</text></error></iq>

em, yeah nobody does it like that

Oh because the fields are there?

Works with Prosody

<iq id="supported_query_fields" to="some_room@conference.localhost" type="get" xmlns="jabber:client"><query xmlns="urn:xmpp:mam:2"/></iq>

correct

then i would say you dont respect the XEP zash

the xep says server must only use the default if the field is omitted

not if the value is omitted

and it returns: <iq xmlns="jabber:client" xml:lang="en" to="admin@localhost/106417850532246324662946" from="some_room@conference.localhost" type="result" id="supported_query_fields"> <query xmlns="urn:xmpp:mam:2"> <x xmlns="jabber❌data" type="form"> <field var="FORM_TYPE" type="hidden"> <value>urn:xmpp:mam:2</value> </field> <field var="with" type="jid-single" label="User JID" /> <field var="start" type="text-single" label="Search from the date" /> <field var="end" type="text-single" label="Search until the date" /> <field var="withtext" type="text-single" label="Search the text" /> </x> </query> </iq>

leosbrf, you need to do a "set" query, not a "get"

what you request is the filter options

but you dont need filters

the xep is a bit weird this way, but there are 2 important distinctions

No code I have can distinguish between <field> being omitted and <value> being omitted.

filterng and limiting

as you see the filtering options dont help you with your usecase

so just set no filter, which would tirgger the server to give you all messages in the archive

but because you dont want all messages, you limit this result now to one message with a specific id

you can do this with adding a rsm node

lovetox , how can I limit that? What is the stanza that I need to send to server?

<set xmlns='http://jabber.org/protocol/rsm'> <max>1</max> <after>09af3-cc343-b409f</after> </set>

but as pointed out, you would need the message id, before the one you want

you sadly cant request a specific id, you can only get all message after an id or before an id

https://xmpp.org/extensions/xep-0313.html#query-paging

btw what xep do you want to implement here that needs that?

I stuck in getting that id. Because when I request the room history, the stanzas returned have this format: <message xmlns="jabber:client" to="admin@localhost/3247206387247491563010" from="some_room@conference.localhost"> <result xmlns="urn:xmpp:mam:2" id="1578509255305907"> <forwarded xmlns="urn:xmpp:forward:0"> <message xmlns="jabber:client" xml:lang="en" from="some_room@conference.localhost/leo1" type="groupchat"> <x xmlns="http://jabber.org/protocol/muc#user"> <item jid="leo1@localhost/107114335247801189862018" /> </x> <archived xmlns="urn:xmpp:mam:tmp" by="some_room@conference.localhost" id="1578509255305907" /> <stanza-id xmlns="urn:xmpp:sid:0" by="some_room@conference.localhost" id="1578509255305907" /> <urls /> <body>someone send a message</body> </message> <delay xmlns="urn:xmpp:delay" from="conference.localhost" stamp="2020-01-08T18:47:35.305907Z" /> </forwarded> </result> </message>

what xep you want to implement?

what is "reply to"

what is "reply to"? that is a similar functionality that whatsapp provides (you click a message and click reply). The way I implement, I would store the message id in the child message

lovetox 👆

so not a XEP

you doing your own thing

yeah attaching something to a message is a big discussion right now, and its currently not possible

because MAM is not adapted to that usecase

hmm, so that is the reason why I spent some hours researching and found none implementation..hehe . But something you said before gave me directions. In this example: <after>09af3-cc343-b409f</after>. Is this code something dynamically generated for the current query request?

@lovetox , Anyway, thank you for the information provided