XSF Discussion - 2019-03-06

can someone point me where in the RFCs it's said that a client must ignore 'to' of incoming stanzas?

I might be blind, cannot find it

ah, found

zinid, where?

I only had https://tools.ietf.org/html/rfc6120#section-8.1.1.1 in mind

yes, it is

there is no need for the client to check the 'to' address of incoming stanzas. However, if the client does check the 'to' address then it is suggested to check at most the bare JID portion (not the full JID), since the 'to' address might be the user's bare JID, the client's current full JID, or even a full JID with a different resourcepart

that is not really "most ignore", isn't it? ;)

sorry for not being exactly precise in this bureacratic chatroom!

I'm currently challenged by "check whether the user has finished entering a JID yet". The current naive version just lets the server hammer DNS and s2s...

https://op-co.de/tmp/yaxim-jid-search.mp4

Ge0rG: I don't think I've ever had say an email client attempt that… why check for valid JIDs that way?

waqas: I hope for better usability. First I was going to filter based on the public suffix list, but then I realized it's huge, complicated and messy, and it doesn't support local deployments

"it doesn't support local deployments" — wat? browsers do embed it, and tend to work locally...

waqas: I mean it won't accept "mycoolxmpp.local"

If you want local stuff, well "helloworld" is a valid local hostname

Also I'd have to either embed an immediately outdated list or load it at runtime periodically.

waqas: that's exactly my point

I'm struggling to see the usability goal. What are you trying to solve for?

People mistyping hostname?

Giving people a way to explore xmpp? Dunno

I wonder if there's a use-case for a directory of all publicly accessible XMPP servers

It would be a good thing for auto completion.

> Giving people a way to explore xmpp? That's not needed

> Giving people a way to explore MQTT?

For me the demo looked good 👍 I'd use that.

Ge0rG, maybe add an explicit search button and prepare to integrate it with Jabber Search

jonas’: it'a already integrated, you just need to open the lower group

I'm still not sure about "have all ways to add JIDs in one dialog" vs "omg what is all this shit"

Jabber Search is not muclumbus

Wait. What's Jabber Search then?

XEP something

something something.

I don’t know it off the top of my head, you’ll find it in the XEP list, I’m at work

https://op-co.de/tmp/smack-typesystem.png

*sigh*

Ge0rG, you asked: http://upload.zinid.ru/xep-eax-csr.html

Is slixmpp now the only maintained fork of sleek?

Kev, are there any others?

That's what I was asking.

No you're asking if slixmpp is the only one :p

I know only of slixmpp

Link Mauve, ^

That’s the only one I know of.

And I’d really like to merge both projects, the only thing SleekXMPP has got for it is its python2 support, and that is getting EoL’d quick.

SleekXMPP still seems pretty popular, so if we could just reuse the name and the website and rename the next slixmpp version that, it’d be perfect.

Ge0rG, https://github.com/Flowdalic/android-xmpp-iot-demo/blob/master/app/src/main/java/com/clayster/xmppiotdemo/JidTextWatcher.java#L61 is what I did a while ago

regarding user provided jid string verification

flow: yeah, but I also need it to work on domain JIDs, and this is somewhat harder ;)

flow: did you see last night's yaxim video?

Ge0rG, no

flow: what I'm essentially looking for is "does the whole string / what comes after an @ qualify as a domain part"

https://op-co.de/tmp/yaxim-jid-search.mp4

flow: ^

Ge0rG: can't you just cheat somehow? I dont know... Fire dns srv queries on the client?... :)

Wiktor: my ISP is already serving ServFail to SRV requests... with a 3s delay.

Sending a disco#info is actually the least work for me.

Dns over HTTPS! For every problem I've got two solutions a.k.a "next problems"

for faster access better grab a local client index of all jabber domains, as TLS certs are now logged in Certificate Transparency logs the list is there... but it would take a lot of space... so use bloom filters, see, easy! ;-)

no, I just got a better, CaaS - Completion as a Service, you send a query and it autocompletes the domain part, then you can start offering it to others for a small fee... it's an unmet need right there :)

Ge0rG, the "what comes after the @ qualify as domain part" check is what the code I linked does

plus, does the string before the @ qualify as localpart

flow: it only checks whether what's between @ and / is non-empty.

Ge0rG, I don't think so

Wiktor: https://foaas.com/greed/cloud/Ge0rG 😁

Wow, great... 👍

flow: it calls JidUtil.isValidEntityBareJid() which splits and calls XmppStringUtils.parseDomain() which just splits, and then checks the length.

flow: unless you added magic in smack 4.4

or rather, jxmpp 0.6.4

Ge0rG, hmm, no?

you possibly missed that line: https://github.com/igniterealtime/jxmpp/blob/0.6/jxmpp-jid/src/main/java/org/jxmpp/jid/util/JidUtil.java#L140

flow: ah, right, it goes down into Domainpart.from() which also ensures domainprep.

but I'm not sure how much that actually adds.

flow: or, in other words, does it add any value over attempting to JidCreate.from(string)?

btw, great twitter (uh!) thread what people thinking about XMPP: https://twitter.com/bascule/status/1102693889842921472 😀

Ge0rG, it does. It ensures that there is a "valid" localpart and domainpart, and no resourcepart, and provides better feedback about what is missing for the string to become a valid JID

flow: but I'm hiding the error anyway, at least in the last version ;)

https://twitter.com/DLitz/status/1102961301616156676 🤣🤣🤣

Ge0rG, wut?

I actually replied to this

Has anyone got a server-side implementation of MIX-PAM (XEP-0405) anywhere?

zinid: yeah, saw that a bit later. Master trolling at work.

dwd, I have

dwd, part of ejabberd 19.02 actually

Damn android single tasking. If I open Twitter links from yaxim, then use yaxim, the Twitter client forgets what I was looking at.

Ge0rG, yeah, that's annoying AF

Ge0rG, and btw I didn't mean to troll, but whatever

zinid: not speaking of you! 😀

oh, really, I'm impressed

yeah, that dude just goes full strawman hoping from subject to subject

and gains a ton of likes because, well, XMPP is godawful

zinid, Oh, really? That's awesome. I'll check it out.

dwd, I can create you an account at my server, I'm running mix pam there

zinid, Is it a module? And is it available (open source, downloadable, whatever)?

dwd, yes, it's a module (actually two) and is a part of ejabberd distribution

While that thread is actually rather sad, it's showing Twitter at its best. You can try to convince them, or try to insult them, but it's not going to change anything.

zinid, Do you have any clients or MIX servers that use it?

Oh, wow. Found a client library that supports MIX, and claims to have a GUI client. (Which is an example only - I've not attempted to see if it has a MIX UI yet).

dwd, slixmpp has OMEMO, Jingle, and MAM support (I think?), poezio has none of that :P

dwd: no I don't, Daniel has some experimental branch of Conversations though, we tested our implementations

zinid, https://github.com/qxmpp-project/qxmpp might work, I have no clue.

pep., Yeah, the MAM is handy, and the '60, but no MIX. I might have to fix that.

Ge0rG: true, they have already set up their minds, and I am not sure this is actually attributed to Twitter

Heh, qxmpp, is that what the telepathy people starting to use?

for telepathy-nonsense?

Ge0rG: this is just blatant IT hype and groupthinking

zinid: twitter is known for this kind of toxicity. It's also a good place for short, "state-less" rants making it impossible to have a proper debate based on merit

it's like XMPP! 😁

Ge0rG: well you can debate at other more appropriate place, but you can do nothing when you're told that XML is shit (which is I agree BTW)

and every xmpp discussion ends up there

zinid: I agree with that too, and I'd love to see them solving extensibility on top of JSON :>

But they can't.

Ge0rG: you can do that in asn.1 😁

oids ftw!

Anyway, with Matrix we do have the anti-design to XMPP. Monolithic spec instead of modular, JSON instead of XML, persistent message database instead of a stream of ephemeral events. Let's wait five years and see what sticks.

Extensibility is easy in JSON. Permissionless extensibility is somewhat harder.

zinid, mmm, OIDs, which you can only allocate with a bureaucratic process which doesn’t even work?

yeah, sticks where?

jonas’: why doesn't it work? last time we registered oids in iana

in 2004 😂

But my usual argument about XMPP and XML is that we don't use XML like SOAP - instead, it's very terse XML and pragmatically-driven, so while it's still XML, it's vastly nicer than what they may have experienced before.

> persistent message database instead of a stream of ephemeral events. That one is actually good

Sorta - you can record a stream, but you can't make a persistent database ephemeral.

dwd: yeah, but I'd love to see those arguments spelled at public, not in this chat where everyone is already converted 😁

zinid, For sure.

zinid, Also, the "Nobody uses XMPP" thing, except for half the games industry, military, government, and so on.

also, though xml is not, let's say, ideal, it's not as fatal as opponents trying to draw it ✏

it's just not ideal, indeed

Indeed. There are larger problems in XMPP. Maybe we should tell them about the mess that is MUC, Carbons and message routing (which are just my favorite pet problems)

But but it's confusing! 😂

> That one is actually good Andrew Nenakhov: agreed

pep., qxmpp is what Kaidan is using if I recall correctly, lnj?

Ge0rG: we can say that of course, but they won't even understand those problem may exist, because designing protocols is easy

you just need to stick to json and http!

> That one is actually good although seems like the design is awful, looking at how their servers behaves

zinid: implement matrix protocol in ejabberd! 😁😁

and you probably cannot make such type of replication terribly efficient, so matrix will be always slower than xmpp by design

Ge0rG: we considered that a few years ago and I had the discussion about it a few *days* ago with the boss

the conclusion is "this will require too much effort, let's wait if matrix demonstrate some real progress except hype"

5 years have passed and where is the progress?

What's 'this'?

Zash: implementing matrix in ejabberd

note that after 5 years of evolving of xmpp we had several servers and clients and rfcs

Implementing the c2s protocol is probably doable, then we can steal all their clients. ;)

The s2s protocol .. yeah, no.

Zash: doable of course, but requires full time commitment and we are currently limited in human resources at p1

s/ at p1//

mqtt and sip were much easier to implement

also, the only attractive part of the protocol is replication, everything else is terrible, they reiterate the mistakes of xmpp

and I personally rather disappointed in federation and looking at P2P

Doesn't MQTT provide way less functionality than what XMPP-Core does?

It seems weird to me to compare MQTT to XMPP

Zash: sure, mqtt is very dumb

Zash: you cannot compare them

I don't understand why people do that all the time

I personally don't think that full p2p will have great UX

Andrew Nenakhov: not full

I could say here wait for me to complete my xor proposal, but seems like nobody reads them 😂

Zash: did you read the matrix s2s spec BTW? Is it finalized?

zinid: Not really and not really.

ok, let's wait 5 more years

> Also, the "Nobody uses XMPP" thing, except for half the games industry, military, government, and so on. dwd: oh, just saw it. yeah, virtually every second our customer is gamedev, hehe. xmpp is quite strong there.

and again, also the reason I consider MIX: group chats in gamedev is hell, thousands of participants

Our Groupchats should probably handle that well and quite easy to implement

Andrew Nenakhov: 😉

zinid: it would be really great to have testimonials from them about XMPP

zinid: or maybe at least contact people to interview, cc winfried

Ge0rG: hum, let me explain how their development is done: they hire some developers (usually from India) and assign them to work with us. their project managers are almost clueless and their bosses are totally clueless. so the question is whom to ask? that developer from India?

chat infrastructure is actually a huge pain for gamedev they offload the work because that's not their area of competence

zinid: good point. Is there some kind of in-house senior tech project manager?

so it's just like to ask you to promote that cool Japan engine in your murican car.

Ge0rG: I don't contact them, so I am also not the right person to ask, sorry

and thanks gods I don't 😁

Haha!

> pep., qxmpp is what Kaidan is using if I recall correctly, lnj? Yes, that's right and I'm working on the MIX implementation in QXmpp. It's not completely finished though and most of the pull requests haven't been merged yet.

zinid, I assigned XEP-0415 to XOR and XEP-0416 to EAX

push is done, waiting for the build to happen, so ETA 1h or so

jonas’, thanks a lot

https://rosenzweig.io/blog/the-federation-fallacy.html

zinid, #old, I linked that yesterday :p

It was here a couple of days ago

argh

sorry guys, I'm too slow

https://lwn.net/Articles/781205/ comments

oh, good start: they attacked the author's graphs!

Don't read the comments

is it Dirac, power-law or loglog distribution? hum... so many questions

That funny article failed to address the most famous federated protocol: email

No it did

You probably missed some points

stop throwing links! I need to work still.

The part about how gmail has 100% of the market and self-hosting is impossible?

Ge0rG, I just copied it! You should have read it already

pep., It didn't provide any graphs for email, I mean

Zash, I'm not saying I agree with all she's saying

zinid: it's right behind your CA proto-XEP, but it will move up if you keep talking about it

Ge0rG, I shut up, carry on

nah, just joking, I cannot shut up

Just claim that 'most people have email on Gmail', but what about outlook, Yahoo, mail.ru, hundreds of them noticeable big servers

Andrew Nenakhov, still power-law distribution

Yes, but: not sharp and with plenty other choices.

I already have brought an article somewhere from 2001 where the same thesis about federation was spelled

zinid, I'm interested if you can find it again, but don't bother if it's too much of a hassle

they are all basically the same "federation doesn't solve all problems" ok, agreed, also the sky is blue and water is wet, why the long blog post?

Funny thing is, what do they offer as an alternative?

Wikipedia

?

Centralized services, 'because federations are centralized anyway'?

Facebook!

pep., http://shirky.com/writings/powerlaw_weblog.html

pep., from 2003, sorry!

> In systems where many people are free to choose between many options, a small subset of the whole will get a disproportionate amount of traffic (or attention, or income), even if no members of the system actively work towards such an outcome. The very act of choosing, spread widely enough and freely enough, creates a power law distribution

thanks

so predicted 20 years ago

Andrew Nenakhov, centralization is a bad solution obviously

it has the same drawbacks plus others

That reminds me to search for ...

and find http://urbanpolicy.berkeley.edu/pdf/AQWJEG0804PB.pdf

btw, today I noticed that my xmpp domain was suspended (I forgot to pay for it). It's becoming quite annoying to self-host even if you have the needed qualification.

That has happened so many times to me. >_<

The DNS Mafia!

zinid: it always was an extortion business, but with the buy your own TLD for $$$ business, it became unbearable.

We need an alternative DNS root!

I propose the XSF.

right after the CA!

after me please

Also DNS-over-XMPP!

we wanted to buy process.one and guess what? It's $8000 per year. Tell me about CA mafia now 😛

oh, it's still free, no surprise

Ge0rG, why not cloudflare does it literally every other possible way https://developers.cloudflare.com/1.1.1.1/fun-stuff/

it would avoid the TLS setup cost, should be pretty performant actually

moparisthebest: that list is missing XMPP.

Ge0rG, add that to the missing integrations page :p

I wonder if SamWhited knows somebody who can be convinced to make it happen

it should be pretty easy to make a component to do that, hmmmm

great I needed another useless project

If you want I have a whole list I can share with you

the list of useless things?

If you can do some of these for me

I already have a whole list of my own thanks

zinid, useless and not useless. Often the line is pretty thin :P

if I create a xmpp component to answer DNS queries, and add support to a DNS proxy to use it, will anyone in here use it ? :P

moparisthebest: Aware of any DNS client libraries that let you easily feed them data without it coming from the network?

If so, give me BASE64'd DNSSEC-signed blobs plz!

But then, what problem are you solving?

I was not aware of one so I wrote one https://github.com/moparisthebest/jDnsProxy it would be easy to, using that, write a xmpp component to answer them, and a resolver to ask said xmpp component :)

same problem DNS-over-HTTPS and DNS-over-TLS is solving I guess?

maybe faster though, if the S2S stream is already established

Please make a bot that runs `dig $message_content` and responds with the stdout+stderr content.

:D

I promise not to do evil. Not too much at least.

17:24:55 Vlatin Pudimir> !dig SRV _xmpp-server._tcp.muc.xmpp.org 17:24:56 foorl> _xmpp-server._tcp.muc.xmpp.org (SRV): 1 1 9269 xmpp.xmpp.org.

moparisthebest: It won't help clients do initial connections tho, where we have actual problems still.

jonas’: now all we need to do is to host that bot as a component on 1.1.1.1

I'm sure Ge0rG can tell us (or already did?) how many fail at SRV lookups

can you have SRV records for IP addresses?

Zash, it will if the resolvers are set up to use it, like my router is

Routers that don't support SRV lookups but they'll support DNS-over-XMPP?

Wow, my ISP fixed their resolver. It's not servfailing after 3s any more.

but wire-format-wise it'd just be like DOH, base64'd raw TCP request bytes

Zash, yea honestly endpoints will end up using DOH etc first before all the routers get fixed, android phones already do DNS-over-TLS to google by default now

firefox also can or does

Dns over xmpp sounds like a perfect April 1st XEP folks

better than jackbeer?

how about this, if I can write the code to get a working POC in the next, 2 weeks max, will someone else commit to writing the XEP :D Wiktor ? :)

moparisthebest, the XEP for April 1st you mean?

yea

not sure if trolling

ah

moparisthebest: are you going to map all DNS payloads to XML elements?

hell no, down that path only lies tears

tears of joy!

it'll be just like DOH, base64 raw DNS packet bytes

😂

moparisthebest: 👍

call it DOX ?

moparisthebest: that's not XMPP style.

Then we get cloud flare in the loop

it's not HTTP style either yet here we are

moparisthebest: if you do that, you can as well just run jonas’' SRV bot.

They'd like the April joke I bet haha

Ge0rG, no it'd be actually useful, because you'd have a UDP/TCP DNS port locally you can query, put in resolv.conf or whatever, except it's just connected to an XMPP server to resolve names

for, some definition of useful :P

wait a minute, Ge0rG was complaining about his ISP failing DNS queries... if only he had DNS-over-XMPP!

he soon will!

There's already DNS over HTTPS and HTTP over XMPP, so if you chain them together, then you can easily get DNS over XMPP

so... what's the latest/greatest java xmpp library for clients nowadays...

uhoreg: that sounds like there is no need for an own XEP

moparisthebest: sed

oh, I finally finished reading that "federation sucks" article

great ending: > To see the future of liberty for the many, look not to the obscurity of XMPP, for arcane technical voodoo can never be wielded by those who need it most. To see the free future, look to Wikipedia.

re council@: I think it is a terrible idea if we start to omit namespace bumps on experimental XEPs. That will only lead to confusion and a bad reputation wrt. XMPP's interoperability

flow, so, after the publication of my XEPs I now need to bump the namespaces after every second PR? nowaaaay

however, not that I'm against that much

zinid, that I why I am in favor of an "incubating" state of XEPs, where they got an stable identifier and a clearly visibile notice that there may be no namespace bumps on backwards incompatible changes

flow, if you incubate too much how they will become visible?

flow: that's called "Experimental"

Ge0rG, I hope not

whatever, I'm not against at bumping the namespace at every commit, I know sed!

I see people starting to implement XEPs and then wondering why their implementation does not interoperate with another one

Namespace bumping is one of the worst ways to iterate protocols.

Ge0rG, I think you mean backwards incompatible changes are the worst way to iterate protocols

flow, but when you implement IETF I-D you don't wonder?

Unless it is a major rewrite, protocols should come with feature tags.

but sometimes they can not be avoided, but in this case you need a namespace bump

at least for my definition of backwards incompatible change

zinid, exactly

probably because an ID has not RFC number

flow, I kinda agree that we need a concept of a "living standard" and "final"

everything in between should be burnt with fire

but an experimental XEP has an XEP number

I honestly prefer numbers, it's easier to reference them

it may appear as a small difference but I think it has an huge impact on how people perceive the standard

zinid, as soon as the thing is out of the incubating state, give it a number

XEP-DRAFT-0123

DER-0123!

or, even much better, a short name

I never understood why we use numbers when humans are much better with letters

flow: let's namespace-bump MUC and see what happens?

I already memorized the short names of zinids new XEPs, and have a idea what is behind every name

I doubt I would have been able to memorize it if they where assigned a number on submission

flow, my next xep will have a terrible name, I cannot invent it

Should the upcoming sprint be announced on xmpp.org ?

Yes

summit, fosdem, sprint