XSF Discussion - 2017-11-24

Ge0rG, and everyone else, https://wiki.xmpp.org/web/index.php?title=Easy_Group_Chats&curid=12&diff=9796&oldid=18

MattJ: nobody has stolen your approach, you don't have it implemented

Link Mauve, +1

I still have no good idea how to glue all the layers together in my client to actaully be able to implement that chain of fallbacks :/

my thoughts exactly when I try to implement new cool xmpp things: calling hooks in unexpected places is annoying

for example, dealing with messages forks and carbons is a nightmare

zinid, you should involve yourself in the Message Routing 2.0 discussion, if you aren’t already

things might in fact get simpler

I always wonder whether this perceived complexity is from my inexperience with 1Mloc+ projects or if this is specific to the IM domain or if this is specific to the XMPP domain

I think that's because of paradigms change

you mean existing code having hardship with adapting to the new situations?

we used to have simple chatting where you don't need to *synchronize*

no, I mean IM paradigm

now we face the problem which is in fact data *synchronization*

take a look at mam and carbons

those are clearly replica transfers

sure

so I would suggest to implement this replica layer first

but I have no idea how to make the transition :/

I also think XML is not the best way to exchange replicas, lol

anyway, we can invent new and shiny IM 2.0, but how to make the transition?

that's why I virtually don't discuss the topic

I have no ideas

zinid, there have been very concrete proposals

jonasw: I didn't find the solution with ancient servers

I mean I didn't see such proposal

(using Bind 2.0 as a signal for "use IM 2.0 rules" and some s2s protocol to make serves negotiate that between them, PLUS and here’s where your ancient servers come into play, fallback to the existing Hints for Carbons and MAM by writing them explicitly into the messages when transferring to IM 1.0 domains)

fuck yeah, much easier :D

much easier than what?

[10:26:12] <jonasw> things might in fact get simpler

what did you mean here? :)

I see only more complexity, it looks for me new servers need to maintain old and new behaviour

zinid, inside the IM 2.0 domain, you’d see much simpler rules for archiving and carbons than what we currently have

probably, but those rules still suck in the sense of replica synchronization

do they?

for example eventual consistency is not guaranteed at all

isn’t it? ✏

you can lose messages easily

how?

via s2s failures?

stream management?

it doesn't solve all the problems

sure, but many

no

it doesn't solve a problem with server outage for significant time

and also, the current discussion is mostly about different clients on the same account

do you think it should be extended to different accounts?

typically the message is buffered for 5 minutes

on server outage with significant time I’d expect message loss, it’s as simple as that

yes, I know what about the current discussion ;)

the sender would get an error back, it’s a rather clear situation?

jonasw: well, I have quite some experiences in database design and I would say consistency should be guaranteed even with huge outage

jonasw: yes, it will receive error back, then what? the recipient loses the message forever

it is clear to the sender that the recipient has never seen the message

or that the message was never even sent

yeah, that's not how eventual consistency works ;)

I find it consistent.

both nodes end up to "know" the "message was not sent" state

you face inconsistent state and you need to resolve it *eventually*

(the sender explicitly by seeing the error, the recipient implicitly by not seeing the message)

no, recipient node might know jack shit

due to unexpected failures

this is acceptable in p2p computing for example

and is tolerated

but not in xmpp ;)

distributed databases even have anti-entropy technics to deal with rotten bits (https://en.wikipedia.org/wiki/Software_rot), but let's say we don't need this in xmpp (at least so far)

I find matrix approach very decent in this regard, however their use of http polling and json is annoying

Exactly.

yeah, the idea is great, implementation is shit

I still think servers should not replicate, it should be done peer-to-peer

servers could maintain replicas, but this is not required

The replicas are just caches in Matrix I think. At least I don't see why servers couldn't just ditch and later re-fetch them.

probably, I don't know all the details

but look, this is much easier to fetch replicas p2p: when you become available you ask for replicas of your contacts (possessing the last seen message id)

you don't even need stream management queues for this

except that you wouldn’t want your phone to drain its battery on syncing other peoples messages

jonasw: yes, so replication servers are needed, so they can reply to such replica requests by themselves (without client involvement)

at which point you’re ditching the p2p approach and we’re kind-of back to "ask for MAM on your server"

no, I can disable this feature on my server

but this can be implemented via MAM, yes

we already have this for groupchats

why contacts are different?

see above?

what exactly do you mean? MAM should be considered as just another replica or cache

I mean the power consumption issue.

of course there can be tons of replicas, so?

what issue if you ask your server to maintain replicas?

I don’t see how that would work, discovery and security-wise.

I don’t have any issue with that, it is simply MAM

I don't find it less secure than current approach

anyways, gotta go, I already typed too much today (hand injury) :(

jonasw: this is kinda "enchanced" mam :)

“07:54:06 jonasw> on server outage with significant time I’d expect message loss, it’s as simple as that”, people coming from email would be sad to hear that, it’s been a solved problem for a long time there.

Link Mauve, when you have a long outage, emails will bounce just like IM messages do

the term "long" is relative and may be different for IM and Email

It's *Instant* Messaging after all

Zash: nah, it's data replication now, that's the problem

Yeah :(

and we're trying to solve this replication problem via IM

Hence the huge whiteboard I’ve been promised for years. :(

FOSDEM?

Yes! Brussels 2018 - The Whiteboardening

sounds like I really need to sort that out

Let’s use Markdown! https://statuscode.ch/2017/11/from-markdown-to-rce-in-atom/

ah, they failed to sanitize HTML :)

the HTML support in the markdown component is left enabled intentionally

html inside markdown looks like a really insane idea ;)

zinid, markdown is specified as an HTML superset, IIRC

markdown is defined as a html superset

too slow

wat?

a html superset?

yup

what does that mean?

every valid HTML document is a valid markdown document

oh shi~

(even though I’m not sure that’s actually true, there are probably some corner cases) ✏

but it’s intended to be that way

actually, every text is a valid markdown document, no?

yes

so markdown is a text superset, okay

but the semantics of a valid html document interpreted as markdown are the same of the html docmuent interpreted as html

nevertheless, the article is interesting

I wonder what those german guys asking about XEP-0392 are about.

jonasw, what do you mean? that's the jsxc developer

the client that is mostly known for its integration into nextcloud

daniel, and the other three?

*two

christian schudt is the babbler developer

(not saying that it’s a bad thing or something, I was just wondering why suddenly)

you know people, it’s scary

and klaus herberth is i believe also jsxc

yeah, Klaus Herberth has jsxc in his mail adress, so I assumed you were talking about him when referring to "the jsxc developer"

I see. I was originially talking about marcel because that's the first email i saw

ah ok

Fwiw I don't know any of these people. I just know of them. Babbler is super awesome.

jonasw, https://ppjet.bouah.net/poezio-colors.png one use case of why using the bare JID for colors is meh

nick would be better in that case. These are all IRC users, going through a bridge (Gluttony)

pep., jonasw already said it on the list

Ah I have to catch up

too many germans in jabber

says the russian

too many humans

mathieui: russians in jabber can only write spam bots and talk shit, mostly ;)

daniel, do you want a MAM namespace bump?

Flow: no?!

or, in other words, isn't a MAM namespace bump required for your proposed excluded-types?

or is the service required to return an error on unknown fields? which would be good I think

Flow, i think the additional filter thingy is off the table anyway

we just gonna adjust the rules

ohh :( I like the idea of additional flexible filters which are discoverable

Flow, custom filters are already part of the protocol

and not specified in xep313

what about making MAM responses type=headline?

Aaaand changing the default configuration to "N/A"

daniel, right, but there is currently no specification for message type filters

Ge0rG, I can't follow, but it's likely because it has been a long week

I believe that MAM archives should be allowed to also archive groupchat messages FWIW

if they were specified they wouldn't be custom, no?

Flow: type=headline would better reflect the ephemeral status of MAM responses, and the rationale for "N/A" is in https://mail.jabber.org/pipermail/standards/2017-November/033762.html (c)

Flow: why do you believe that way?

daniel, probably, but does not matter, I want a mandatory set of filter mechanism which can be extended by discoverable additional filters, so that we can experiment which filters are a good idea and which are unnecessary

Ge0rG, am i reading this mail correctly you are planning to come to the summit?

Ge0rG, most for the same reasons that Kev said

Flow: I disagree with Kev on that.

daniel: I can't promise that.

I really can't.

Ge0rG, My vision is to give MAM services maximal freedom regarding what to archive, and clients should be given the filtering mechanisms they need

Flow: what about using a separate XEP for all of that?

Ge0rG, for the f

iltering mechs?

(the optinal filtering mechs)

What gets stored should be up to server policy. What gets can be described by a XEP

Flow: no, for things that go beyond normal im client message archival.

Zash: implementation defied!!!!

Ge0rG, wouldn't that be an MAM alternative?

Flow: what about using the same backend storage but not return everything to normal MAM requests?

Ge0rG, if you still have the possiblity to return groupchat and probably all other types of messages, then that would be fine in my book

err, to return if the client asked for it

Flow: the server is free to store anything it wants, as longs as it only returns what's written in MAM on a normal MAM query

Flow, it's not my XEP; i could live with the filtering on query proposal. (also I believe storing groupchat in the personal archive has downsides that make it not really fit for use). I believe existing implementations would ignore unknown form fields. thus introducing new ones could be done without a NS bump

so to answer you question no I'm not planning a NS bump. I'm not planning anything

daniel, guess my initial question was if you think that your proposal 2) would require a NS bump

I'm not sure if the service ignoring unknown fields means that you can avoid a namespace bump

really, let's get rid of namespace bumps altogether.

i generally tend to be on the rather not bump if it doesn't cause issues with existing implementations faction

Yeah for unexpected behavior

I'm sure it's possible to replace almost every ns bump with a feature.

Ge0rG, well then it's ok

but in this specific case, I don't want to extend MAM.

but existing implementations need to change anway because they a) need to start actutally storing groupchats b) add a database field for the type to make the query efficent

are there provisions in MAM to full-body-search for messages?

Ge0rG, I don't think so

Flow: so what's your use case again?

And while at it, I want to point out that K9 did for years client local searches. It took a while until they implemented server side IMAP searches

Flow: that's actually an argument against gc in mam

No, it's just an observation regarding the importance of a server side search feature

Flow, maybe make your argument for groupchat in the user archive on the list

yes please

not sure if the relevant people are following this discussion

in here

seems like the list = this room

all the same people around

zinid, i'm not sure people read the backlog here. i certainly don't

zinid: there is a formal difference.

zinid: Council needs to catch up on standards@

daniel: I'm reading backlog when I have nothing to do ;)

daniel: actually I rant here only when I'm bored :)

zinid: then it looks like you need a {boy,girl,other}friend

Ge0rG: nah thanks, I'm good with my wife and son

when is that lottery drawing?

which was it again

euro something?

Guus: https://www.euro-millions.com/results

tx

no results yet

41 minutes?

ah, 21 CET (that's in an hour)

Now I wonder if somebody can hack that web page.

what the fuck is this?

zinid, could you be more specific and less explicit please?

crap, crying kid, afk

Guus: no

other questions?

It wasn't a question, it was a request: please act in a civil manner here - it makes for a better atmosphere, which helps us to achieve more.

Guus: I act as I want, stop teaching me, I really don't care how you think one should act

zinid: I'm sorry to hear that.

Yeah, people have different views, deal with it 😀

Really, if you cannot tolerate words like crap or shit you're better off not using internet

zinid: this place is not made for your crap and shit, but for constructive discussions in the context of XMPP. There are plenty of other places on the internet where you can behave like an a*hole and make friends that way...

Ge0rG: oh, I started to think you will not chime in 😂

zinid: I've got a decade worth of channel operator experience in Germany's worst IRCNet channel, I have a sixth sense for trolling time... 😜

In other news: the Euro-millions draw is in progress!

animated gifs and all!

I'll give calculation a shot, but it'd be good to have someone else verify my results.

(although technically, Alex will probably sign-off anyway?)

Ge0rG: so euro millions is a constructive discussion?

"in the context of XMPP"

zinid: ah, that's were you were getting at earlier. Yes, it is, actually.

the board member vote were in a tie. The tie-breaker is a procedure that takes the outcome off this euro millions draw.

I didn't get anything

tie, producer, draw, wtf

ah, procedure

"that's where you were getting at" <-- I now understand the confusion.

I'm so excited, much wow

most background is here: https://mail.jabber.org/pipermail/members/2017-November/008695.html

not interested, I don't care about your policy and your invented rules

I'm interested in standards

suit yourself. We can definately use your help with the standards!

whoever you will elect, jabber will continue degrading, I'm 99% sure

Guus: I don't write documents, too much effort, 95% of time I will spend with english-russian dictionary

or russian-english, whatever ;)

Don't worry, there are lots of ways you can help out with the standards that don't require writing stuff down.

this looks like a sect

zinid: people are having fun with technical protocols to ensure fairness of an election. If you compare that to elections in the USA, where districts are rebuilt to optimize vote results, or your country, where the president gets elected with 115%, it's a really great process... 😝

Nah. We're all friends. Here, have a sip of my kool-aid.

Ge0rG: not sure what technical protocols you mean. like your election bot, lol?

Guus: that's what Jim Jones said...

zinid: if you weren't so busy cursing, you'd already know

yeah, if I was interested

but I don't give a shit, luckily ;)

zinid: you are bored anyway, give it a try

Results are in! By my calcultion: 19 - 24 - 28 - 30 - 50 -> 19.24.28.30.50./ -> f2507b627edc74d540e7cb747138ed7ee8e19d4458f9db5bfd3a92b5c97a9541 -> Nÿco

zinid: if you can't be bothered to behave like a normal person, please leave. I'm serious.

ralphm: who are you?

I'm the XSF Chair

wow

what if I tell no?

I'm not sure what your thing is, but I've been getting complaints. We like this community to be inclusive and I don't appreciate our hospitality to be abused. By asking such a question, it appears that you don't care. If I'm wrong about this, please let me know.

yeah, I already understood that you build only the community you like, people who you like to see, without any technical merits

so how far did this policy bring you?

xmpp is a degrading protocol which is nobody want to implement and even here about

Feel free to leave if you think we are doomed

nah, I like here

Well, good. We welcome discussions on technical matters, in a civilized matter. It may all be funny for you to repeat how bad you think the protocol, our procedures, or the people in our community are. I don't find it acceptable. If you insist on keeping that up, I will simply remove you from the XSF maintained venues like this room.

ralphm: why don't you remove yourself? I don't even know you, what did you do for xmpp? I don't remember you discussing the protocol. Why do you have power to remove developers from this room?

I'd you can't be bothered to even look that up, that's on you. This room is hosted by the XMPP Standards Foundation. I'm its Chairman. It's pretty much similar to how I would throw out a guest in my house if he starts bothering the other guests.

ralphm: ah, so you're a boss here

If you need that kind of confirmation, yes.

ralphm: so your task is to ban developers you don't like, boss?

zinid: yes. Bye.

ralphm: good job!

I had one job, I guess.

some things the world can do without

I can only hope he fairs better in other settings, so banning him from the world seemed excessive. For now.

Fares, even. Yay auto-correct.

To bad it came to this. Always feels like a loss to me.

Not sure what the provocation was about.

It doesn't take much provocation to make him try to insult you. It would be more of a loss if new people see that kind of abuse and don't get involved because he'll just call them an idiot.

I've had multiple complaints, and I'd like to think I'm a reasonable and generally nice guy. I also read back some previous discussions. Before, other people asked nicely. I did a final attempt. He keeps on trolling. Fine.

I've suggested improvements to other people before, and those people have changed their ways. I can sometimes be annoyed, but this is simply intentionally riling up people, and I don't think we need to tolerate this.

Oh, I didn't want to to imply you were unreasonable.

Good. You are actually close-by.

Close-by?

Physically close. As in within a few kilometers.

Sorry I didn't come by last weekend. How was t-dose?

ralphm: uneventful, sadly.

Right. They have never seemed to live up to their goal of being a bit like FOSDEM-in-the-fall. Too bad.

Ugh, WTF.

I don't think "what the fuck is this?" was even remotely meant to be a provocation.

I mean I can't speak for zinid but this quite obviously reads like "what on earth is this Euro-millions thing?" to me.

And I'm not sure he called an individual an idiot.

If it was just that, it would be just fine. I say things like that all the time

He's called multiple people idiots and been asked to stop, though he might not have used those words this time.

Aha.

Holger: I only asked him to be civil, and stopped when he made clear he had no intention.

I've made a point of not escalate things. Sadly, that did not work.

I see him using flippant slang. As for the substance, I see him ranting against how this community does things. I totally understand how he won't make friends that way.

I'm perfectly fine with arguments and dont mind expletives at all - but I do think it's important to have civil discourse.

But I won't agree with this as a reason for a ban.

He was asked to stop multiple times by different people including Guus, me, George, and probably others and only responded with abuse. Then someone who actually has authority politely warned him and he still didn't listen, I'm not sure what else there is to do besides a ban.

If he tells a new person that their spec or ideas are garbage they're not likely to come back; no community needs that kind of toxic behavior, no matter how smart the person is.

Holger: just read back the logs for a few days. He's more than welcome if he can behave like a normal person.

I'm going for some sleep now.

Goodnight

SamWhited: Yes, different people use different tone, and asking them to change it tends to not work that well in my experience.

How many days of backlog am I supposed to read to find the reason for the ban? I skimmed through November now. I found some purely technical comments on various details, some ranting on MIX, EXI, Twitter, Java, the XSF membership/council election process, and our priorities. I guess I overlooked something.

Whatever, this is your show, I'm not even a member. I'll shut up.