XSF Discussion - 2019-01-07

Hi. I don't get Evgeny answer, can anybody explain me how clustered database are affected by the fact that we use or not an attribute to specify level of ordering?

goffi: Most distributed and replicated databases tend to be eventually consistent, such that one db node may have a different modification timestamp for an item than another node in the cluster.

(mysql and postgres for example have async replication, they are ACID for a node, but not for the cluster; many nosql dbs are similar)

waqas: so he's not actually answering to the question (which is about adding an attribute to handle the order vs using order of <order/> elements), but talking about the feature in general, and only arguing about the modification time right?

And this sounds like implementation details, nothing prevent to add a field to the database which is set when the item is received, before any replication.

"order by" is used in SQL which can be used with clusters, so I don't see how this is different here.

goffi: Consistent order for a mutable ordering value like last-modified time (vs an "immutable one", like creation-time) really works in SQL dbs with a single node, or which have cluster-wide write locks.

In mysql for example, a write to one db replica does may take some time before it's written to the rest of the cluster.

So different nodes in the cluster may have a different understanding of last-modified time at any given point in time

Pagination is tough as well. You request page 1. Then you request page 2, except page 1 results may already be invalid by the time you ask for page 2.

Since last modified order is changing all the time

I don't want to speak for Evgeny, I'm just assuming the above is what he means.

Database synchronization is hard.

waqas: pagination issue is already mentioned in RSM. I'm not against adding a flag in the result saying it may be inconsistent.

goffi, no pagination is not covered with RSM as soon as you order by a mutable field

well, it is covered

but it’s not safe anymore

"The set of items that match a query MAY change over time, even during the time that a requesting entity pages through the result set (e.g., a set of chatrooms, since rooms can be created and destroyed at any time). The paging protocol outlined in this section is designed so that entities MAY provide the following features" (§2.2)

I'm confused by your one email which says " the order affect the whole archive, not just a result set" and in a following email "it's true that the order inside the page is not influenced."

waqas: thanks about the explanation. I think this should be mentioned in the XEP, and maybe a flag could be added.

MattJ: the first message was because yesterday it didn't seemed clear that the ordering was on the full history (i.e. we are not ordering the says 10 items we ask for, but we are ordering the whole archive and getting the 10 first items from that)

Right

Matthew: the second is because there is currently no mechanism to say, inside the page result we get, that we want the first item to be the most recent or the least recent. Currently the XEP state (not explictly enough apparently, I'll fix that), that the first item of the page is always the most recent, regarless if we are getting the result normally, or backward using RSM.

I'm personally leaning towards just having another field in the existing MAM query form, to keep things simple

and then we can fix pubsub separately

waqas: by the way, the last modification time is currently the default ordering using XEP-0060 without anything else. There is actually no "modification" notion, it's just a new item, but that the result it the same.

I think the problems are slightly different, because you can't update a MAM entry (and never will be able to)

MattJ: you mean extending XEP-0313 directly or havint a new XEP which extend MAM ?

I was planning to extend XEP-0313

MattJ: Message corrections...

and the problem with MAM isn't which property to sort on, it's simply which direction to sort in

The problem with pubsub is not the direction, but the property

> I think the problems are slightly different, because you can't update a MAM entry (and never will be able to) Not 100% true, because of redaction, but close enough.

and another reason not to solve both with the same protocol, is that the server implementation for both of these is quite different

MattJ: what about MAM with Pubsub (which is the case I'm most interested in) ? The item can be overwritten there.

Yet the order-by XEP only allows discovery of support for <order>, not within a specific protocol

cant we just implement SQL replication over XMPP and be done?

I can fix that and only use is with MAM for Pubsub, it doesn't really make sense in the other case anyway.

I don't know - I have nothing to do with MAM-with-pubsub, never have, and I'm still unconvinced whether I ever will

I think filtering should have been a pubsub thing instead of a MAM thing, but today it's the only tool we have to filter items in Pubsub.

and it's a really important feature for most of use cases I have with Pubsub.

the MAM example in the order-by protoXEP is about Pubsub

if we add order reversal to MAM, interaction with order-by would get confusing if that also let you select asc/desc

Otherwise, they could probably live together

But I don't know if I can imagine implementing it personally

Hmm

Maybe I can, in the context of pubsub, but basically it would be a separate implementation

PubSub in Prosody is squeezed through an abstract LRU cache, so that level would have to be thrown out and rebuilt

goffi, does a config item on the pubsub node definitely not fulfil your requirements?

(a config item that says "sort by update" or "sort by creation")

"keep order on ower-write"

or do you need this control at the time of query?

There's no update in PubSub, only implicit deletion

I think the basic use-case being supported is just showing microblog posts by time they were first posted vs time when they were last modified. Last modified is current state, right?

MattJ: for some use cases I need to do it at query time. For instance for a forum, I want to be able to have last modified message popup on top, or have it by creation date.

Zash: no there is no update in Pubsub, that's why the "date of modification" is explained in the XEP, the business logic is changed.

waqas: last modified is current state yes.

The idea is to _add_ business logic so that overwriting implies updating, right?

ralphm: yes

But also to have a way, later, to order by a field for instance (e.g. prioriy of a ticket, or evaluation of an answer to a question).

Hi folks.

Hey dwd, back from holidays?

Yeah, holidays plus family emergency, unfortunately.

dwd: sorry to hear that. Hope things are normalizing again

So I have an interesting problem with the XMPP Summit, and I wondered if anyone else had the same issue.

My new boss said he looked at the details for the XMPP Summit, and concluded it wasn't happening because there was nothing on the website and wiki page.

the wiki page doesn’t look exactly empty tom e?

dwd - I am happy to facilitate later tonight

I'm about on my way back home, which will take some time.

In the mean time, if someone feels like drafting a blog post... 🙂

we do have a blog post, don’t we?

jonas’, The Agenda is empty, for instance.

https://xmpp.org/2018/11/xmpp-summit-23/

jonas’, And the blog post *is* there, but several articles down and not easily visible from the website.

(that’s true, which is why I didn’t link it in the first place)

so, some kind of banner on the home page?

the summit is members-only, isn’t it?

maybe linking the blog post from the wiki would be a start?

pin it

jonas’, Not as far as I know.

ok, I need to go now

Guus, I'll catch you later...

dwd, if what you say is true, a banner on the website would make sense, indeed

but I would welcome more exposure

I can probably draft some HTML for a b anner on the website, but someone will have to do the design fine-tuning

jonas’: what about a box of upcoming events?

Ge0rG, complexity

I want this simple now

it could be on the wiki as well, and feature elections and the like

speaking of which... there is still no election page for Q1

> I can probably draft some HTML for a b anner on the website, but someone will have to do the design fine-tuning 💓

that’s a nice box you got there

jonas’, The problem isn't just a banner. The problem is that if it weren't for my insisting it's really A Thing, my manager wouldn't have believed it's a real event.

dwd, yes, I see your point

however, the banner is something I can do, and which I realised when browsing $anotherOrgs page recently that it is something the XSF lacks

The banner, linking to a Thing-definion

dwd: isn't that kind of symptomatic for XSF events?

(announcing the event loudly on the main page)

SCAM needs a bigger marketing budget?

Ge0rG, Yes.

But here's the thing - right now, XMPP jobs, whilst relatively few, are very highly paid. There seems to be a lot of XMPP about, in some well-resourced areas. We should be getting really good attendence.

So I'm wondering if the reason we're not is because it's not seen as a normal corporate-friendly event.

Cobol jobs have the same classification

dwd, for starters, from my perception, summit was members-only

we’re not really advertising it to non-members

jonas’, No, and that can't be a good thing.

from my perception, Summit is heavily biased to new protocol design, which is not typically something for newcomers

jonas’, The Dinner *is* members-only, though we've rarely enforced that.

The one time I attended The Dinner, I wasn't a member.

Ge0rG, Yes, but we had an interesting session abotu where XMPP was used last time, which was really fascinating to me.

Members +1 ish?

Zash, Roughly.

But I wonder if we could [re-]introduce talks and other things that look a little more conference-y. We've done so before, but I think that making the Summit look a little more professional (ie, corporate) would help people expense travel to it, and therefore attend.

seems legit

dwd: that sounds like a great idea. Maybe have one day of talks and one day of workshops?

Even just calling what we do every year "workshops" will already make it more enterprisey

not that I consider myself to have a chance to attend on company budget, considering that we’re so not XMPP

That sounds fine

Let's add this to the agenda of the upcoming summit?

Also, I would not mind having more than one per year

there used to be two, across the pond

We're we could differentiate

Ge0rG, I was thinking just an afternoon to start with, but yes.

dwd: would you mind drafting a semi static summit event description for on the website?

dwd: an afternoon of talks? Sure

We can have jonas’ banner link to it

There's a start.

There's these sprint things

Guus, I'll scribble some text.

Zash, You mean the hackathon type things? The IETF seems to get people to come to those.

Then add the sprint thing (cc pep. )

ohmygod

`npm i` then run `grunt` to compile Sass.

why

I take my "I can do a banner" back

Ok, I'm starting my car engine, afk.

dwd: something something ask pep.

jonas’, This is why I said I would "scribble some text".

jonas’: what in the what and also why?

!

https://github.com/xsf/xmpp.org/pull/498 dwd, Guus

I think running a Real Conference for XMPP might be a good idea.

Yes, and that doesn't need to be the Summit

Doit

Kev, I agree, but I think adapting the Summit to increase engagement probably gives the most overall benefit.

I don't think "We've got a room in the Cisco offices for a couple of days" is really it, though. What we do is perfect (well, YKWIM) for what we get out of it, which is advancing the state of XMPP with technical discussions, but a long way off being appealing as a Conference.

Which isn't to say we couldn't do a thing instead of the Summit (in it's usual form) instead, at the same time

Well, if we re-cast the bulk of the Summit as being "workshops" and "open panels", it might work, but I take your point.

It might, but I don't think one room at Cisco is what we need for this.

There's nothing stopping us doing a Conference and having protocol advancement sessions, either.

except manpower

But probably not all of it, and I think it needs more prior organisation than a Summit.

Those are separate goals - make Summit look more enterprisey while keeping the content / make a conference appealing to commercial xmpp users

There's "Try to make the Summit easy to expense for the usual sort of crowd" and "draw more people to the Summit"

> There's "Try to make the Summit easy to expense for the usual sort of crowd" and "draw more people to the Summit" Does that mean that newbies should be able to participate and understand most of the "things" there?

Depends on what kind of newbie

Like me

So usually there is a bunch of discussion on protocol, like the last few times with MIX.

Some of those can be really indepth, some not so.

What kind of participant would such an event aim for?

So it depends on whether newbie means: “I haven't been around for long”, or “I don't know much about the protocols”

I also want to note that we potentially have more rooms this year. Rudy (of Cisco) wrote: “I did in the meanwhile book the rooms Cassiopea and Electra (can be combined to a single larger, wider space) as well as the adjacent room Centaurus which you could use as a break room for speakers, organisers, logistics, ….”

ralphm: Where is Cisco located

Diegem

https://wiki.xmpp.org/web/Summit_23

https://www.openstreetmap.org/way/29455013

> So it depends on whether newbie means: “I haven't been around for long”, or “I don't know much about the protocols” I dont know much about the protocolls. My fear is that just the time and this MUC arent enough to get out of this kind of newbie status...

what about https://xmpp.org/extensions/inbox/jingle-xtls.html

how long does stuff stay in the inbox?

forever?

until the end of time

mrDoctorWho

lovetox, it’s never deleted

https://dpaste.de/BjkD/raw something like that for sprints would do? thoughts?

I'll send a patch

Sure!

https://github.com/xsf/xmpp.org/pull/499

I'm back home

Thanks for all of the ideas, people

wb

I'd suggest keeping the summit pretty much as-is, and have another, more conferency, event, at another date.

I simply like the summit format - it has its own value.

I tend to agree, even though I only participated remotely

side note: the Cisco offices are not publicly accessible.

(I’ll have to figure out how to do that this year :/)

for a conferency-type of deal, we might want to consider using a different venue

agreed

I'm all for this new idea :)

well, maybe not new idea

But I want to see it implemented

In any case, I'd suggest to not deviate to much from the Summit format for the upcoming event - as it's close, I'd like to avoid confusion.

very much +1

Sure

btw, I’m happy to hack in more crappy inline CSS to make this fancier: https://github.com/xsf/xmpp.org/pull/498

but I’m at the limits of my creativity

so if anyone provides a mockup, I’m happy to try to achieve that

jonas’ it looks great to me - although I'd prefer a flashy / contrasting color, to pull in attention

nit: can this be vcentered?

the text inside the box

and cente....whathesaid

centering is trivial (and you probably mean hcenter)

I mean vcenter

please suggest colors, I’m really bad at picking those

pep., oh, you’re right

reddish or yellowish

Guus, I guessed that much

I actually ment vcenter 🙂

the details are the tricky part

vcentering was still one of the fun things to do when ran away from "the web"

in this case, I just had to kill the margin-bottom of the p which holds the text

jonas’ we use this blue for active menus: maybe that'd work: #008CBA

makes the link unreadable obviously

I could make the entire thing clickable though

which would do the trick

yeah, I suck at this more than you do 😃

https://sotecware.net/images/dont-puush-me/OSJcUNBVjbMmS8ipFMSXmQUiLGu7U6qA7XjNO1t-3lQ.png

I think this is fair

LGTM

updated the PR: https://github.com/xsf/xmpp.org/pull/498

❤

If I get it well, I find this comment deeply depressing: https://news.ycombinator.com/item?id=18847472

Telehash is a project from Jeremie Miller right?

well, it's an answer to a direct question on the status of telehash: https://news.ycombinator.com/item?id=18827032

so it's not an unsolicited "by the way, look at this" type of comment.

Guus: yes, and it's coming from the Telehash core team apparently.

Guus: no that's not my point, my point is that Telehash, a project founded by Jeremie Miller, creator of Jabber, is using Slack to communicate.

looking at the screenshot after a while, this looks *so* much more alive now

with the upcoming event and all

we should be doing this for all our upcoming events

@jonas` definitely.

I wonder whether we can somehow integrate this with the blog, so that we don’t have to micro-manage tihs, but I don’t think that’s easy

jonas’, sprints and all?

yes

yes the banner looks nice.

i'd love to replace that top-right wiki thing

top-right wiki thing?

goffi: temas was one of the folks running jabberstudio.org back in the early days :-|

fippo: that's even more depressing :(

jonas’ https://wiki.xmpp.org/web/Main_Page, top right, "upcoming events"

ahh

it needs an update, too

I’m going ahead and deleting the past stuff

I'd love for the website to become more 'active' or 'community driven', or whatver

unlikely to happen with the current framework

we need to figure out why it doesn’t build correctly locally

jonas’ please do, but move them to the archive page.

which archive page? ✏

see link in box below 'recent events' 🙂

what I typically do is move 'upcoming' to 'recent', and move the last few 'recents' to the archive.

ugh

I’m incompetent sometimes

also, doing like three things at the same time while having had way not enough sleep isn’t good for me

fixing

hehe

oh cool, Evgeny message was a misunderstanding.

also, it might be good to mention that the bar for adding/modifying the website is a lot lower than some people might think. I invite everyone to suggest changes.

Guus, does that look good to you? https://wiki.xmpp.org/web/Main_Page

jonas’ sure, thanks.

it's no biggy either way

regarding the website bar: it is non-trivial to test the website locally due to the strict version requirements on pelican

not to mention that CSS changes require some (possibly web "standard", but still) npm installation stuff

right, it might currently be hard to apply a change, tehcnically, but I had a discussion last week where people where under the impression that only a few people get to decide what goes on the website

effectively, only a few people do ;-)

tohse with power over the uhge green merge button

right, but I'd _love_ to change that

I don’t

sure, there are a few gatekeepers, but content-wise, I'd love it live more.

yes

PRs are excellent

for this.

yes

If a gate-keeper could merge or review my PR for the sprints that'd be great :)

pep. I actually added a comment 🙂

I second Guus’ comment

heh, sure I can change that. So I'd reverse the order of the list as well right

(also, Seve , I'm painfully aware that there's still your PR to review)

what’s blocking that?

my time.

or attention span

simply didn't get around to it.

I mean, is this a content or a technical issue which needs to be reviewed?

note that one cannot test the menu locally with current pelican versions as of now

Should I add dates in these lists?

pep., dates would be good, especially for upcoming

pep. if you think it's helpful to interested people, yeah!

try to answer where, when, and if possible, what

jonas’ I simply haven't taken the time to look at it yet.

What is usually a bit difficult, as it's discussed really close to the date

Except for Brussels maybe where it was said it would be about UX from the start

pep. do something like: "medio January"

"or Q1 2019"

I sense a misunderstanding

hmm?

misunderstanding?

Is it me or there's a lot of latin being used in this community? :P

Guus, aren’t the sprint dates fixed?

like, known ✏

They are

goddamnit

Well, at some point

so I think that Guus misread your sentence "What is usually a bit difficult, as it's discussed really close to the date" (which I did too)

"What" is usually a bit difficult

yeah

https://bpaste.net/raw/57647708077f ?

Not sure how that will show

Aaah, I was under the impression that the final date was known only pretty close to the event.

so, a misunderstanding indeed. 🙂

pep., so, yeah, exact dates please

this was just confusion

pep, if you have actual dates (days), I'd use those.

yup, sorry for that.

hmm, brussels is two not consecutive days, in different months

That's going to be annoying

hehe, whatever works 🙂 I'll not twist your arm behind your back to have exact dates 🙂

I'm already very happy that this goes on the website in the first place

Thanks, I appreciate

I like my arm very much where it is ✏

pep., what

I don’t get it

if it’s two different events, it should get two entries

so you can put exact dates on each?

It's the "same" event, just that the summit "appeared" in between :P

sorry, I think I’m too tired to follow this

I think roel wanted to do two days, but asking people to be there tuesday _and_ wednesday was a bit too much, considering most of us are going to the summit and fosdem.

So they're going to do 30th (wed), and there's something that might happen on the 1st, at fosdem maybe, I don't know much

I see

pep. just add wednesday, and note on the wiki page that there might be a follow-up after FOSDEM

I’d just list "Jan 30th, Feb 1st, 2019 (around FOSDEM), Brussels" as link text

*At fosdem

or just Jan 30th, that’d probably work too

yeah

I'd be happy with any of the above. Go for it 🙂

Should I use british or usican on the website? ✏

(tired)

so tired

go for US when in doubt

we use it in XEPs, too

use letters for months

and avoid confusion

yeah

or ISO format, but words read better in the list

"March 30-31st, 2019." ?

I'd not use the year

maybe in the archive

I’d add the year

you can never be sure whether a website is kept up-to-date

but for the upcoming events, they should be sufficiently close

I would add the year as well, it can be confusing otherwise

and when reading a list, it’s always good to see that right away

and yeah what jonas’ says

pep., that format looks readable to me

I’m worried about the "still" :)

:D

got to keep the suspense goin' 🙂

:D

https://github.com/xsf/xmpp.org/pull/499

Was dwd going to write a text descirbing the summit?

I’m not sure

pep., "March" is now duplicated in the list item

Ah fail.

I’d also probably have put it this way: * [Brussels, January 30th, 2019](...) (around FOSDEM)

but that’s just my preference

not going to twist arms, either

Please stop threatening my arms

I am not threatening your (or anybodies) arms :)

time to get creative with the body parts?

not the trout!

:D

(t)rusty, probably

here, pushed again

also, isn’t trusty EOL ;-)

unrelated side note: https://github.com/xsf/xmpp.org/issues/497

this concerns me

I like it now, pep. ✏

Guus, it concerns me, too

board should definitely look into that

I would’ve made a PR

push the button, jonas’ !

but I’m not sure where we list sponsors, typically

and how

and such

youknow

Guus, as board member, can you maybe comment that the Board will take this seriously?

leaving this uncommented for 3 days is... a bit meh

you're right

I HIT THE BUTTON

:)

(I am tired and getting silly)

How long does the website take to build again?

ages.

Guus, no

in total I think 10 - 15 minutes until it’s live

k, thanks

that's what I said.

🙂

Guus, you’re not an editor, obviously

hahaha

that takes 60 minutes+

touche

so ... the website is rather fast.

we have a super-fancy banner on https://xmpp.org now :) thanks for the input :)

jonas’, we need the full package, incremental builds etc. :P

Or do you have that already

no.

then it would take like 5 minutes, not 60

well, while we're being effective: is there anything that we can do to speed https://github.com/xmpp-observatory along?

somebody needs to figure out how to teach the involved DNS relsover libraries about the new root zone keys

that’s the current remaining issue with xmpp.net

https://xmpp.org/community/events.html "Krakow (soon)", any idea when this "soon" is due for?

(that I am aware of)

pep., remove it

(or maybe check the git-blame first)

Also same page, it references Summit 22

jonas’ I'm not even sure I understand what you said.

pep., clean it, while you’re at it

pep. also, that's the link to last years' summit

Guus, xmpp.net works, except for the issue that it cannot validate DNSSEC currently. So any zone which uses DNSSEC appears to be broken to it. This needs fixing.

Guus, yes

you can safely remove all events, they're from last year, at best.

(it cannot validate DNSSEC because it uses the old keys for the root zone which have been exchanged recently in a surprisingly ostentatious ceremony)

Guus, so somebody needs to teach the libraries which are baked into the xmppoke docker thing about the new keys

I don't know how DNS zone / key / ceremony works.

it is fun

is this a configuration thingy, or a code change thingy?

probably code change thingy

because the keys are embedded in the library

worst case one needs to patch the binary file

https://github.com/xsf/xmpp.org/pull/500

pep, please remove all events

they're from 2017

oh

oh, actually

no

the berlin one was from last december

The berlin meetup is ongoing

I mean it's recurring, rather

stockholm can be removed

FOSDEM can also be updated

london can be removed

and Paris is also outdated

so, only keep Berlin 🙂

pushed

looks good. I'll let Travis do its thing, but am happy to merge

Thanks for doing all this guys!

goffi, https://wiki.xmpp.org/web/FOSDEM_2019 do you mind adding your talk if you have time?

There's also JC in there

Dele I guess

pep.: yes I will, but probably tomorrow, start to be tired now

thanks

jonas’ are you technically capable of doing the DNS zone magic that you just referred to, or should I direct my pretty-pleases at MattJ or others?

pep.: please ping me again if I haven't done tomorrow evening

k

Guus, direct at me

But I have less than zero time at the moment

MattJ I have cherries-on-top to offer, to sweeten the deal.

ah

is there anyone else capable of doing this?

with zero+ time available, hopefully?

MattJ, does that mean Guus owes you time now

For asking

It's just a matter of updating the Dockerfile to install a newer version of a dependency from a package that Zash already made

pep. I have a tab going with him 🙂

and modifying it to stop using that dependency as a submodule which it currently compiles itself iirc

oh, I can do that, I think

MattJ / Zash can you identify said dependency, and the replacement package?

luaunbound?

(sounds DNSsie?)

yes

I mean, it is dnssie

I'm guestimating that this is the desired fix: https://code.zash.se/luaunbound/rev/3c3f017f35ac

lovetox: why are you interested in XTLS? client-to-client as in Jingle or something else?

yes jingle, but im not interested in it, Gajim already implements it

im just wondering why it never made it experimental

lovetox: encrypted jingle?

ok, yes, it's in the code.

https://github.com/xmpp-observatory/xmppoke/pull/3 <-- incredibly naive copy/paste thingy

jonas’ how do I test this fix locally? I need to run against a zone with DNS SEC?

Hmm, I'm getting 'bogus dnssec' banners, which probably is an indication of failure. 😕

Guus: yes, that's the bug

Pretty sure I confirmed a fix in a temporary container where I installed the package manually

MattJ: im unsure what I'm doing. Locally, it didn't appear to work. Maybe I need to provide the commit message suggested trust file somehow.