jdev - 2019-09-05

...the moment you realize that you need to also add an <x/> element to MUC-PM delivery receipts.

they *are* MUC-PMs

I considered that obvious :)

https://lab.louiz.org/poezio/slixmpp/blob/master/slixmpp/plugins/xep_0059/rsm.py#L28 why would amount default to 10? `max` in the XEP is not defaulting to anything is it?

I guess this specific implementation expects a value, because later on it does `self.query[self.interface]['rsm']['max'] = str(self.amount)`, and everybody knows that `str(None)` is of course `"None"`

just to confirm i'm not missing anything from reading '48, '60 and '163... there's no standardised way of storing/retrieving bookmarks individually instead of as a list of them, right? even when using pubsub you're expected to treat the list of bookmarks as a single item ("current"), or am i wrong?

Correct

But now, introducing Bookmarks 2: https://xmpp.org/extensions/xep-0402.html

thanks Zash, I didn't thought it was really serious given that title, actually; will give it a read

But then you get the problem that you can only have one bookmark with older PEP implementations.

Probably want all these to just be backed by the same store serverside

If only we didn't have to cope with legacy systems.

i can consider myself lucky; it's for a custom implementation in a non-federated service so...

I think very often we don't, and our determination to make life hard for ourselves by doing so is a hindrance.

Kev: it is a hindrance, but my gut feeling is that by breaking backward compatibility, we will drive ourselfs irrelevant even faster

Backwards compatibility is one thing.

We often try to ensure *forward* compatibility, and that's a world of pain.

We often also overengineer our XEPs

I'm not sure to what extent that's true. We often try to make big generic solutions, and then years later find people are using them in interesting and unexpected ways. 60'd be a good example of a big complex thing that's seeing lots of use because of it's genericness (not in IM so much, but in other interesting systems).

In other times when we've tried to go for simple, pragmatic solutions to get things working, we've found dealing with the edge cases has been unpleasant (45, 308, etc.).

I guess protocol design is hard, maybe?

0045 is not really simple, it comes with a huge number of unspecified corner cases.

That's kinda what Kev said, right?

I'd consider 0308 as almost elegant, except for three minor points: the number of old messages that may be corrected, the limitation to the full JID (that's gone now?), and the method to reference messages, which is a mess of its own and not the fault of 0308

I'm saying that 0045 is the opposite of simple, pragmatic.

Hi I created openfire(v4.2.3) server on my virtual machine( ubuntu). When I tried using spark client 2.8.3 to login, I go this error '“certificate can’t be authenticated”. I got a similar error when I used pidgin. However, pidgin automatically corrected and I’m able to connect using pidgin. The thing is, I'm trying to connect using node-xmpp-client and I get this error: XMPP authentication error. What am I doing wrong ?

jid: bent3n @localhost

FQDN: localhost

database: postgreSQL

What am I doing wrong ?

tom, so, what I did for that weather stuff is that clients request weather forecasts as they need it.

the XMPP <-> Weather API proxy caches and reacts to rate-limiting HTTP status codes appropriately.

I didn’t do polling and then publishing into pubsub, because the clients are often idle for several hours, not needing any updates, so that would simply save resources on all ends of the equation

Also, back when I started that project, I wasn’t comfortable with PubSub yet :)

is that the only reason you didn't use pubsub? If you had to do it all over again would you?

Where did that come from? /me scans archives

I ask because meteorology is one of my interests

Zash, operators@

tom, I’m not sure, I think for the forecasts it doesn’t make sense to use PubSub still, at least not a normal PubSub service

and I've done projects like that before, although my problems were more about getting the sensor data from the sensors to a server in zero-infrastructure areas

if anything, the XMPP<->HTTP translator would have to be a PubSub service on its own

so that clients can subscribe for updates based on their location

I've been working on a dynamic self-healing mesh network

publishing all weather forecasts for all possible locations to pubsub doesn’t really make sense ✏

but the thing is, I still need something to buffer the scientific data until the server can be reached to ingest it

I'm thinking because of the reliable nature of XMPP, I could just use an XMPP library and send stanzas with it and have the library buffer those in case it can't reach the server

and when it does flush the buffer

> reliable nature of XMPP

I think that is a dangerous statement unless you really know what you’re doing

well what should I know?

you need to know about stream management, and you should be using IQs for data transfer

be able to deal with duplicate deliveries

that’s probably already a good start

and have a plan for when the server is longer unreachable than you can buffer

that's what I'm talking about

also, what you’re describing sounds like a Delay/Disruption Tolerant Network, which is an active field of research in the context of which I actually wrote my Masters thesis

atmospheric conditions change and because the network is dynamic, node all nodes are reachable at all times

another thing I was looking at was A.L.F.R.E.D.

alright fact exchange daemon, but that's not really suited for distributing large facts

and ounce a fact is distributed it overwrites previous facts

tom, maybe have a look at the Bundle Protocol (RFC 5050) and implementations thereof

even if I were to cryptographically sign my facts there's still a DOS vector

(being a mostly academic thing for now to my knowledge, the tooling is rather rough at the edges)

thanks, i'll look into that

What I was thinking so far was to use a fault tolerant layer2 such as the batman advanced routing protocol along with IPv6 multicast groups for distribution and use unicast for re-trans in case a fact was missed due to temporary islanding.

but using XMPP pubsub as a convenient way to access the data casually

but going back to this, there are some interesting XEPs that can make XMPP a little more distributed network-wise

like the server-less zeroconf XMPP configuration where as client can speak directly to each other

zeroconf is an ugly protocol but it does work

jonas’, Can I read your thesis?

you got a website?

I do, but the thesis isn’t on there

which I probably should fix at some point

I’m not sure whether it’ll be useful to you -- it is rather narrowly-focused on implementing a DTN gateway for a sepecific use-case.

(it’s "just" a Masters after all, not a PhD)

it’s not on my universities website, so I guess I have to check with them before sharing it

you don't own the rights to your thesis?

I do

but that doesn’t mean I didn’t grant an exclusive right

although the code is under BSD-3-Clause