On Tue, Jan 13, 2009 at 2:59 PM, Ned Freed
<ned(_dot_)freed(_at_)mrochek(_dot_)com> wrote:
--On Tuesday, January 13, 2009 09:28:27 AM +0100 Arnt Gulbrandsen
<arnt(_at_)gulbrandsen(_dot_)priv(_dot_)no> wrote:
Pushing SCCCP out of the way isn't going to work very well... I suggest
getting a new port (x) and using these rules:
Clients MUST try port x first. If a client gets connection refused on
the
right port number, it MAY try 2000, but in that case it should watch out
for SCCCP. Sieve banners look like <example>, and always contain a line
starting with "SIEVE " and one starting with "IMPLEMENTATION ". SCCCP
banners look like <example>, and always <dan?>.
Instead of standardizing a multi-port search procedure, I suggest we
standardize the use of DNS SRV records to discover managesieve servers.
Excellent idea. I strongly support doing this.
The latest ManageSieve draft already does that ;-).
Sorry, forgot that was in there.
Given that support for SRV records is already required, old servers that are
listening on port 2000 can be made to work with a new client that defaults to a
different port simply by adding an appropriate SRV record. And an old client
that goes straight to port 2000 on the named server isn't going to interoperate
with a new SRV-dependent setup anyway. So I see no compelling reason to
continue to mess around with port 2000. Let's just get a new port assigned, put
it in there and be done with it.
Ned