[lsb-discuss] LSB futures tracker: Rendezvous

Jeff Licquia jeff at licquia.org
Thu Jun 22 13:48:31 PDT 2006


On Thu, 2006-06-22 at 07:15 -0700, Wichmann, Mats D wrote:
> I notice someone created an LSB Futures tracker item
> for Apple's Rendezvous.
> 
> http://www.freestandards.org/futures/candidates/rendezvous/index.html
> 
> Besides the fact that they now call the technology
> Bonjour, I think the problem statement is wrong, it
> shouldn't be "add Rendezvous to the LSB because
> somebody asked for it", I think it should be to add
> automated network service discovery similar to how
> it's provided by Bonjour.  

There are a few levels to this.  There's a network protocol, which is
sometimes called Zeroconf and sometimes called by whatever name Apple is
using for it this week.  That protocol is standardized by the IETF, so
we shouldn't break from that.

Then, there's the question of how Linux developers take advantage of
Zeroconf.  The client perspective isn't too interesting, since we can
either standardize an ABI or expect that people will do the usual socket
stuff on their own.

Where it gets interesting is on the server level, because most Zeroconf
implementations I've seen require you to somehow register your service
with the system before it is advertised by the protocol server.

This may not be all that interesting, since it seems from my quick
observations that Avahi is taking over in the non-Mac Unix space, so we
can just standardize Avahi's registration procedure.  This does, though,
implicitly bring in the CUPS question: to what extent do we want to
force an implementation on non-Linux people in the LSB?

There's also an interesting side-question, which we might want to think
about.  In order to register a protocol, you have to have a protocol ID
of some kind.  This means there must be an ID registry.  How is that
managed?  How can services unique to Linux/Unix get representation in
the registry?  Is that a role the FSG can play?

> So should this tracker be updated, or just thrown
> away and a new one started?

I suppose it depends on how valuable the previous information is.  It
was correct at one time, but isn't really correct any more.  Do we care
about that?




More information about the lsb-discuss mailing list