gnunet-developers
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: gnunet-psyc multicast


From: carlo von lynX
Subject: Re: gnunet-psyc multicast
Date: Tue, 25 Jan 2022 12:04:33 +0100
User-agent: Mutt/1.5.20 (2009-06-14)

Hello Loki!

On Tue, Jan 25, 2022 at 11:55:02AM +0200, Loki Schmidt (They/Them) wrote:
> https://git.gnunet.org/bibliography.git/plain/docs/gnunet-psyc.pdf
> 
> I have always been curious about the multicast implementation of
> PSYC, does it run on top of CADET or is it a parallel service or is
> it intended to extend the unicast functionality of CADET, by
> multicast functionalities.

We tried both ways and got hung up in bugs that we couldn't figure
out. The paper you mention is tg's approach which places multicast
on top of mesh, bypassing CADET. We later tried to start out with
CADET so we have something that works, and optimize into lower layers
later... but then it didn't work with CADET either. I even tried to
integrate GNUnet as a backend into psyced, the historic PSYC server,
as to let the software that already knows parts of the job do its
part on top of CADET - but again, CADET didn't work reliably. I 
heard this has changed, so it would be a good time to go at it again.
My other problem was pragmatic, since psyced.org runs on BSD.

> I have always asked myself it is possible to make CADET have
> multicast functionalities as this would be useful in many
> applications but been unsure if it would be technically possible. As
> running a further network on top of CADET, like currently being done
> with the messenger by TheJackiMonster, sounds wasteful in resources.

It could be possible but one way or another you need to implement
all the enter/leave signaling, controls etc etc somewhere - some of
it needs to be done by the outer application, some needs to go end-
to-end encrypted but some of it needs to be seen and understood by
each node on the way, so it is on different layers of encryption and
routing. Some approaches have been to have these multicast funda-
mentals be part of the binary GNUnet protocol and let PSYC only do
the payloads, whereas other approaches were to use PSYC for signaling
as well, in particular when combined with psyced that expects to be
doing it itself.




reply via email to

[Prev in Thread] Current Thread [Next in Thread]