MURDER or IMAP proxy solution ?

[Rob Siemborski]

On Thu, 1 Jul 2004, Greg Pulfer wrote:

> This made me wonder if I really need the MURDER functionality... First 
> of all can someone exlain me what is a "uniform namespace (no shared 
> mailboxes)", I am not sure if I need that or not. If I could use a IMAP 
> proxy solution and that would make everything simplier I would go for 
> it. Would I acheive the same horizontal scalability with an IMAP proxy 
> solution as with a MURDER solution ?

Basicly, the "uniform namespace" refers to the fact that the aggregator 
allows any client to access any mailbox on the aggregator from a single 
connection to the proxies.  Compare to a simpler solution, like Perdition, 
where once you have authenticated there is basicly a bitpipe to the 
correct backend (or worse -- dns based solution where users have to know 
what server their mailbox is on).

The proxy solution is better because (presumably) the proxy can handle 
pipelining clients better (as in, they don't take a performance hit as 
the proxy examines their commands).  Likewise, it is conceivable that 
getting all of the necessary communications happening within the murder is 
much harder than just pointing a proxy at a database that maps usernames 
to hostnames.

At the time that the document to which you refer was written, it had 
also received substantially more testing, this is not really the case any 
longer.

> Well to explain my case a bit I would like to install a mail 
> architecture for an ISP which currently has approx 200 mailboxes but it 
> will grow and grow and grow, that's why horizontal scalability is 
> needed.

200 mailboxes is nothing.  In your case I'd start with just a single imap 
machine, and decide how to grow later if you need that.  If you start with 
cyrus, its probably about equivilantly easy to move to an aggregator 
system as it is to move to a normal proxy system.

-Rob

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Rob Siemborski * Andrew Systems Group * Cyert Hall 207 * 412-268-7456
Research Systems Programmer * /usr/contributed Gatekeeper

---
Cyrus Home Page: http://asg.web.cmu.edu/cyrus
Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html