[NNTP] LIST EXTENSIONS (again)

Andrew - Supernews andrew at supernews.net
Tue Nov 9 07:16:19 PST 2004


>>>>> "Russ" == Russ Allbery <rra at stanford.edu> writes:

 Russ> What would you think of documenting port 433 for transit
 Russ> connections in the standard and saying that transit feeds
 Russ> SHOULD use port 433 instead of port 119?  That to me seems like
 Russ> the most viable way of eliminating MODE READER.

The problem with this is that it requires changing the behaviour of a
lot of feeder sites. This solution has been around for many years, and
people have been suggesting it should be the standard for many years,
but it's just not getting any traction (and putting it in the spec
won't change that).

We have one (1) peer out of hundreds that's using port 433. Many sites
don't bother to support alternate ports in their own configuration or
administrative scripts, making it harder to get peers if you restrict
yourself to ports other than 119.

 Russ> Redesigning INN so that it can handle reader and transit
 Russ> connections with the same daemon on the same port is something
 Russ> that would essentially require rewriting INN completely.  This
 Russ> isn't just an incidental design decision -- this was one of the
 Russ> core portions of Rich Salz's original design, and it would
 Russ> probably be easier to write a completely new server from
 Russ> scratch than to eliminate INN's split innd/nnrpd structure.

It's also worth noting that other servers' solutions to this either
aren't applicable to INN or aren't even solutions at all. Diablo, for
example, requires that the transit and reader functionality is in
completely separate processes - running both on a single machine
involves considerable contortions to happen at all and doing both on
the same port, INN-style, is simply not possible. Typhoon solves the
problem by using threads. For DNews I don't know the internal details,
but neither do I know anyone getting decent results from running it
on a large feed.

-- 
Andrew, Supernews
http://www.supernews.com




More information about the ietf-nntp mailing list