XSF Discussion - 2018-03-03

Following recent discussions about an alternative OMEMO implementation (licensed under a more permissive license than GPL), I want to inform you, that I am currently revisiting my python implementation and I am very happy with what I'm seeing. My implementation uses pynacl (python bindings to the libsodium library), is licensed under MIT and was written from scratch not depending on libsignal. The only puzzle piece I am missing is XEdDSA, which I have a hard time implementing myself. (I can't even tell you why, the algorithm is not too complex but somehow it just won't work :D) I expect the library to be ready for release about one week after I can get XEdDSA to work.

Syndace: it would be great to have more client developers on board, Gajim and poezio maybe?

Has the XEP moved forward on that issue yet?

Are the implementations out there already using the XEP namespace?

Ge0rG, what do you mean "on board"? As soon as it's released I'm happy about anyone contributing or using the lib, whether the big clients will actually use it is not in my power to decide, but I'll be happy to adjust things to their likings.

Also just to clarify, I'm not doing any xmpp stanza specific stuff, I'm only doing the crypto (and a bit of the wire format).

Syndace: I'm not closely following the development, but AFAICT there are two different OMEMO libraries in python

Ge0rG, AFAIK the other one is just a binding to libsignal and not actively maintained

...and gpl aswell

Syndace: either way, it might be a good thing to plug on the xmpp Newsletter

Oh, there is a newsletter... :D

What's the issue with GPL, it's not like it wouldn't work with gajim or poezio already. The distribution channel would have to be different that's it

Syndace: it's all new!

GPL is a thing of taste, we had the discussion in jdev@ a few days ago. It's more about all the available implementations using libsignal, thus OMEMO can't really move forward.

it's awesome that there will be a non-gpl library soon. however gpl hasn't really stopped omemo thus far. i think by now the number of omemo implementations outnumbers the number of jingle file transfer implementations

pep.: Whiny people who hate the gpl seem to be the only problem with the gpl :)

And yes that, seems people who do the actual work don't care, so just ignore the noise

Counter-example: Swift?

swift doesn't have omemo because of gpl?

Disregarding people who can't use GPL code for legal reasons as "whiny peolpe"..

Zash: can't they simply move to GPL?

It is GPL

The issue was probably more complicated but I don't know all about that

How do you send private messages in a MUC o.O?

That's the first time I saw that :D

depends on your client ;)

maybe you don't ;)

Conversations atm?

I'll just answer for everyone here: I don't have a channel or anything for my OMEMO hacking, it's just something I did on my own when I was bored :D

Syndace: long press the persons avatar enters the private message mode

But if anyone wants to talk about it, feel free to create a small muc or pm me

SamWhited, ah thanks

How can I make something like TTL for xmpp messages?

I think it's pretty rare that a stanza has more than two hops...

MUC

Right

I was about to say. Maybe four if you count muc

Maybe the question is about actual time, not hop count?

-xep amp

Zash: Advanced Message Processing (Standards Track, Draft, 2005-11-30) See: https://xmpp.org/extensions/xep-0079.html

That has expiry, right?

But does it have implementations?

https://xmpp.org/extensions/xep-0079.html#conditions-def-expireat

> Maybe the question is about actual time, not hop count? Yes

Amp is not a good xep if you want to implement self destructible messages

> https://xmpp.org/extensions/xep-0079.html#conditions-def-expireat Does this also work with Clients? I mean do they delet the message after the expiration?

probably not

Of course not.

Only works in proprietary systems

Yes and that's the thing I would have how can a already deliverd message be destroyed on the client?

Just make up your own extension

<clippy> It looks like you want to implement DRM. Please don't.

I lost count on how many times I implemented that

> <clippy> It looks like you want to implement DRM. Please don't. No I just want self destroying messages like in Telegram

that’s DRM :-)

j.r, self-destructing messages don’t really work in a federated system.

they don’t even really work in a non-federated non-proprietary system

but in a federated, non-proprietary system there’s no chance.

It's incompatible with universal turing machines under the control of users.

yeah

Just like DRM

Logical impossibility.

Zash, how about Secure Enclave? :>

jonasw: depends on your definition of work

daniel, "is an actual security device which can be relied upon"

jonasw: I don't know what you are talking about and I probably don't want to.

OK so it isn't possible in Xmpp right?

And don't mention homomorphic computation!

j.r: yes it is.

But not in jabber

Zash, let me ruin your day: that fancy feature in intel processors (SGX) where you can run stuff in isolation from the OS. nice for cloud users, because they can be sure that their stuff is untouched by the provider via remote attestation and whatnot. but in the end it’ll probably be used to make DRM (more) watertight.

> j.r: yes it is. > But not in jabber OK bad to know

you can of course invent something and try to make clients implement it. but if a new (or very old) client doesn’t implement it, tough luck.

not to mention actively evil clients

> you can of course invent something and try to make clients implement it. but if a new (or very old) client doesn’t implement it, tough luck. > not to mention actively evil clients Ok

but an actively evil telegram client or a thing which screencaps the device/whatever would do the same

self-destructing messages is bullshit

You can have an <please-delete-me after="timestamp"/>, but you can't be 100% sure that it'll be respected

If you are ok with that then go for it

yeah, that

Altho that applies to all protocol

> You can have an <please-delete-me after="timestamp"/>, but you can't be 100% sure that it'll be respected > If you are ok with that then go for it But that's possible?

Thanks to the X in XMPP, yes

And with caps, you can know if other clients claim to suppport it

But with MAM and carbons...

you’d have to check that all the peers clients and the peers MAM support it.

And your own MAM

and then hope that they don’t come online after you sending t he message with another client which supports MAM but not your fancy deletion protocol

yeah, that too

and your own other clients

lots of moving parts

So, not really something that can reliably be deployed in an open system

OK thank you

Deeply offtopic, but - if installing a Linux desktop these days, does one bother with swap? 16GB RAM.

Kev, depends on whether you want to be able to do suspend-to-disk

if not, probably not. killing whatever is eating 16 GiB RAM is *probably* better than freezing your system while trying to swap.

Ta.

especially if you do development work. I always try to remember to swapoff -a just in case I create memleak.

*create a memleak

Just put an SSD in my games laptop and thought I might as well dual-boot Ubuntu. So let's see how badly this wrecks booting Windows :)

Swap enabled but swappiness tuned down

(I once had a memleak which filled memory *so* fast that I didn’t realize it was a memleak and thought it was just killing the GPU (some graphics stuff). took me a while to realize...)

Good show.

IIRC the thing is that without swap, you get the OOM killer as soon as memory is filled

yeah, that’s a feature if memory isn’t going to be filled unless by accident

s/by accident/by hip electron app of the month/

Kev: I don't have swap tbh, but I'm not a RAM eater

Kev, okay, add some swap. I am currently at 21 GiB used and I don’t even know how.

much of it seems to be firefox

Too late now, OS is half installed :(

Why is all my RAM used?

ah well

WebKitWebProcess 25..38

And Firefox

Kev, yeah, it’ll probably be fine. and if it isn’t, you can still add a swapfile

Also btrfs. I would need to have a separate partition not as a volume like my rootfs or homefs

I read some long rant that convinced me that running without swap was bad and swappiness=0 was bad

pep.: you should stop whatever you are doing and migrate away from btrfd

*btrfs

Ge0rG, are you going to respond to Kevs mail regarding 401?

marc: the one about using data forms?

Ge0rG, the last one from 28.02

ping

marc: I came to the conclusion that Kev is right and that data forms are the right way to go

Ge0rG: okay, if everybody is fine with it let's merge my two PRs?

marc: yes please. Cc jonasw

marc: yes please. Cc jonasw

marc: yes please. Cc jonasw

Ge0rG: did you write the last message two times?

j.r: is there any protocol where self destructing messages work?

marc: no, it looks like a bug in yaxim

Ie that prevents even something simple like a picture of the screen

Ge0rG: okay, because C shows it two times, Gajim only once

Because you could implement it in xmpp the same way, and it'd work with the same level of security, that is, none

And yaxim shows it three times. 🤦‍♂️

I see three, last one delayed

Out of order too

strange

Could somebody pm me the raw xml please

Oh, looks like a race condition during stream resumption, where sending a message right before <resumed/> will duplicate it

And then it also becomes an "offline message" in yaxim, causing a third transmission

It doesn't actually get sent on the wire before <resumed/> tho?

orwait

Zash: it's sent after <resume>

Between <resume> and <resumed> be dragons.

Yes. Plenty of them

It's nice being the server in that case.

Zash: because you can just queue messages as long as there is a zombie session?

https://petermolnar.net/instant-messenger-hell/