On 5/22/2004 9:07 PM, Gordon Fecyk wrote:
While I've preached small responses and simple yes/no records, I'd like
to know if a small TXT record vs a larger TXT record makes a
difference.
The critical factor here is actually deployment. An RR that is only used
within an enterprise domain can be as big and ugly as it wants, but an RR
that is intended to be used everywhere with all public SMTP transfers
really has to take different factors into consideration.
Assume that the smallest practical lookups are for PTR/A and require 100
bytes in total; a 100-byte MARID RR produces an implementation impact of
doubling the minimum cache (globally). That seems like a silly design
qualifier but there it is. The more lookups folks already do, the smaller
the percentage difference will be, of course.
Where's the limit? I dunno, but storing whole XML documents in the
namespace has got to be the wrong direction.
I'd hope that all those BLs would die off, having served their purpose,
once MARID becomes widespread.
I doubt it very much. The most effective filters I use are connect-level
blacklists, while ~Marid (et al) require the session to get pretty far
underway before tests can be performed (testing MAIL FROM requires at
least that much, while testing the From: header requires full data
transfer, at which point my network resources have already been used).
You've seen the numbers for DMP.
I don't think I have, but if you give me a pointer I'll look.
I only want to know what it does to your numbers. I'm not shooting
down the argument - I'll let the numbers do that or not do that.
Anybody's numbers are just as good. Factor is the important part towards
the cache considerations.
I worry a lot more about the other stuff anyway.
--
Eric A. Hall http://www.ehsco.com/
Internet Core Protocols http://www.oreilly.com/catalog/coreprot/