XSF Discussion - 2019-01-06

More OMEMO!

Nice try :p

OMEMO has a general problem with the key exchange assuming guaranteed message delivery. The key exchange protocol was not designed for lost or out-of-order messages. To mitigate this, clients send key exchange messages until they receive a first encrypted response, which tells them, that the key exchange was successful. So I was wandering, isn't there any mechanism for message delivery that kind of "guarantees" the properties we need?

Or could we use IQs for the initial key exchange instead of messages?

wondering* lol

I see that IQs would take the asynchrony of the protocol, as both parties need to be online for the IQ to succeed.

What level of guarantee?

If it's not 100% then it doesn't really help.

The message must arrive at it's target, otherwise we have to stick to the old hacky workarounds.

If your thing depends on something being perfect then you're going to be dissapointed.

Or at least a mechanism that notifies you about delivery fialure

Something something two generals.

You can stick a counter in there somewhere, and have the other party reply with the last number in sequence they got. Like TCP or such.

this is what OMEMO does, but not for the key exchange, only for the messages after a successful key exchange

Thinking about it again, I guess my search is pointless. Even a notification about failed delivery doesn't fix the issue I have in mind. Only real guaranteed delivery solves the problem and that's something I'm not gonna get.

Well then, dirty hacks it'll be

Push for more xep198

and xep288

What time does the summit usually end on the friday btw, and is there anything planned afterwards

Is there still a need to write minutes for the last meeting?

There is a real comparator of XMPP RFCs and all newer? A lot of devs are lost... - https://tools.ietf.org/html/rfc3920 -> https://tools.ietf.org/html/rfc6120 -> https://tools.ietf.org/html/rfc7590 - https://tools.ietf.org/html/rfc3921 -> https://tools.ietf.org/html/rfc6121 - https://tools.ietf.org/html/rfc4622 -> https://tools.ietf.org/html/rfc5122 - https://tools.ietf.org/html/rfc6122 -> https://tools.ietf.org/html/rfc7622

Anybody seen Alex? It's time to open the Q1 membership season

Link Mauve, that reminds me you have patches for memberbot for slixmpp and python3, we should push that at some point

Also a board(?) question I was kindly asked to forward: is it possible to become an xsf member using a pseudonym?

about the new xep proposed by goffi, do servers even have something like "modification" data implemented right now with pubusb?

are items not just overwritten and then it becomes a new item and not a modified one

also its not clear to me what order by=creation means regarding MAM

items are already ordered by creation, what matters is ASC or DESC

Prosody doesn't, it uses the same mechanism for as for MAM, which is append-only

lovetox: pubsub doesn't has currently something as "modification", it is indeed a new item. The XEP changes the business logic to have this notion, it is explained in §3.

regarding MAM, it's mostly useful for Pubsub, for chat message there is no possibility to overwritte a message, so it's doesn't change anything.

it*

Zash: what happens in Prosody when an item is overwritten, does it appear on top (as it should with XEP-0060 only) or does it stay at the same place?

it deletes the conflicting id and appends the new item

OK

still goffi its not evident for me that saying i want to order something by creation, that it implies ASC or DESC

creation is the column, im ordering my set by

it says nothing about the asc or desc

it gets clearer trough your examples what you want it to mean, but i think this should be spelled out in my opinion

lovetox: yes it does, § 4.1 says "Pubsub service then returns the 3 plays created the most recently, first one being the most recent"

so it's DESC

but I can say it more explicitly if it helps, sure

but please then comment on the mailing list, I'll lost track of stuff in the MUC, on mailing list I can check emails before my next update.

lovetox: ^

i guess it would help, also because you extend this XEP to MAM

do you mean only the pubsub portion of MAM?

or in general?

i guess general

yeah i will write to the mailing list :)

I need it for the pubsub portion, I'm not sure if it's interesting to have it in general, but in the XEP it's general.

I was waiting for feedbacks to be sure if it make sense or not for chat message. For now we can't overwritte anything in MAM archive for chat, so the XEP has no effect (date of creation = date of modification here).

it has an effect if you say creation=DESC because mam sends messages in ASC

ah good point.

but i will add write later a mail with my comments

great, thanks for feedback.

Can you include an 'updated at' timestamp in the payload and order on that in the client?

A delay element maybe?

Atom payloads do have separate published and updated timestamps iirc

Zash: we already have "update at" for blog, but I would have to download all items to order in the client (and the value could be fake)

Ge0rG, There where cases where people where denied membership because the did not reveal the real name in the past

flow: thanks. I haven't seen anything relevant when skimming the application rules

so MAM for pubsub holds the whole history of all items?

so i publish a new one, and then modify it, then MAM has 2 items?

The past is immutable!

so this is a yes

so basically what goffi wants, is sort by creation date DESC then sort by modification date DESC

hm no, he wants the most current blog post but it should not be an edit of a old item

not sure how you can even with that order xep reach that goal

lovetox: for MAM you make a request similar to pubsub, and you can do filtering. It's the pubsub archive, you can overwrite an item.

lovetox: I have already a working implementation (client + server), it's working as expected.

yeah i just want to understand it

do you use now MAM or not?

lovetox: yes, I use it to filter by categories in blogs

lovetox: and I have more uses for it. For instance in tickets, I need to be able to find only "bugs", or only tickets in progress.

but setting an order, has nothing to do with a filter

I'm just talking about my use of MAM here, not the order thing.

how do you filter on message categorys?

Invent a way

I have a custom filter, 'http://salut-a-toi.org/protocols/mam_filter_category' and I specify the category I need.

ok so this has nothing to do with the order xep

no this is to explain how I use MAM

the order is needed in many places thought

For blogs, if I publish "post1", "post2", then "post3", and I find a spelling mistake in "post1", I publish again with this id, but I don't want it to appear on top of my node, like it is the case by default.

But for comments (or forums), I may want to have the last overwritten post first.

and maybe later if I do something similar to stack overflow, I will want to have items ordered by evaluation.

are you aware that server treat every MAM RSM query as its own query to the database

say you want the last 1000 items, but a page size of 10

they dont request 1000 items

they only get 10 and send them to you

That would be an implementation detail of the server

that means every ordering you want, is only executed on these 10 messages

I think in theory you can cache database cursors and try to be smart about them

No the order is on the whole archive, may it's not clear enought in the XEP, but it's about the whole archive, it would make no sense for just a page.

Zash im not saying its a technical limitation, im trying to tell goffi that his XEP does not mention that he wants the order executed on the whole theoretical requested set

I can say that more explicitly, but yes it's about the whole archive.

So, SQL over MAM?

It's not as complexe as SQL, but we have already limit and offset with RSM, order was just the missing part.

I'm planning to read the spec tomorrow, will probably have some feedback ✏

MattJ: cool

OK I'm done for today, .o/

Good night :)

nighty night :)