ietf-asrg
[Top] [All Lists]

Re: [Asrg] SICS

2004-12-25 18:58:33
On Dec 25 2004, gep2(_at_)terabites(_dot_)com wrote:

Maybe, but HEADER analysis might be helpful. (snip)

I fail to see how any of these points address the specific issue being
discussed. A header is part of the DATA, it can be arbitrarily long
and arbitrarily complex to parse. You simply cannot expect to be able to
allocate a fixed amount of resources of all types if you intend to 
read a header (or any number of body lines for that matter).

Moreover, being part of the DATA, if you intend to accept a single
header line, chances are you'll receive the whole message anyway due
to buffering, and what are you going to do with the full message? If
you intend to analyse it, it will be simpler overall to accept a connection
directly through a full SMTP server.

The only way to make the rejection of obvious invalid requests truly
scalable is if you can guarantee a small fixed cost in CPU cycles,
RAM, I/O. Then you have to run this on a dedicated system which
doesn't do unpredictable things like swapping memory pages, spawning
processes, etc. It has to work like a self contained appliance basically, 
only then will you be sure that the resources used are proportional to
the number of incoming requests, ie it's scalable. That also means no
talking to databases, DNSBLs etc.

A mail system with this kind of front end still has to contend with
too much spam, but I think we can then discuss per user spam volume rather
than incoming SMTP bound volume. Making that point was my only aim in
this thread.

Message analysis has its place somewhere after the message is
accepted, but does nothing at all to reduce the SMTP load.

-- 
Laird Breyer.

_______________________________________________
Asrg mailing list
Asrg(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/asrg


<Prev in Thread] Current Thread [Next in Thread>