Re: The messenger service is ready to use

[TheJackiMonster]

I've tested communication in a testbed with multiple peers but with two
devices in local network as well.

It's mostly still a struggle exchanging hello-strings and peer-
identities first to get a connection done but this should improve with
development on TNG and the client-side library which should ease
finding other peers as well.

I'm not currently sure if we should use JSON (which is more common in
current web) or use PSYC (which would be more efficient). The service
itself should allow both, so this would probably be something to decide
on application level.

About secushare: I am probably going to look into the protocol of
secushare again before implementing the client-side library, so there
won't be too many problems making both compatible in the future.
However I can't tell currently if the protocols work together or the
traffic can be handled. I still need stress-tests. ^^'

Also it may be possible to get the client-side library compatible with
Telegram for example but I would first look into getting applications
done with less features. If we have proper interfaces we can think
about compatibility. ^^'

-Jacki

On Sat, 2021-03-06 at 09:46 +0100, carlo von lynX wrote:
> That sounds pretty awesome!!! Does it also work in the wild
> or only within testbed simulations? Does it have any sorts
> of server component or does it manage to distribute state
> in the gnunet?
> 
> The feature list sounds a lot like things that were supposed
> to go into secushare but we didn't get there because we weren't
> able to create stable communication channels between endpoints
> to even start working. Excellent stuff.
> 
> > The library will probably handle more advanced messages based on
> > JSON
> > or some similar text-based format to allow adding features more
> > dynamically. So the service itself does not have to be changed to
> > frequently.
> 
> As the inventor of PSYC I am of course biased, but the rest of
> secushare usually agrees with wanting to employ PSYC instead of
> JSON as it is more efficient in both performance and bandwidth
> and provides for helpfully richer semantics (concepts like
> 'method inheritance' for instance). libpsyc comes with a nice 
> benchmark that shows and explains how and why PSYC performs better.
> 
> The code in GNUnet which has 'psyc' in its name was created in
> a PSYC-oriented spirit, but it doesn't actually implement the
> PSYC wire encoding. libpsyc is the implementation that does the
> equivalent to JSON or XML.
> 
> > Also with JSON messages it should be quite easy getting a good
> > interface done to web clients as well and the messages could
> > include
> > more meta-data for something like posts in social media or forums.
> 
> The implementation could also be Telegram-server compatible so 
> that we can reuse this or that existing client and merely need it
> to connect localhost rather than the builtin IP numbers.
> 
> 
>

signature.asc
Description: This is a digitally signed message part