On Thu, Feb 19, 2004 at 10:29:47AM -0600, wayne wrote:
I decided against VERP and included a cookie header instead; I think it
scales better (multiple MTAs can each add their own cookies) and is less
likely to break things (you don't hit the 64-character limit and you don't
break anything which filters on the envelope sender)
Uh, many (most?) of the bounces that I receive do not contain complete
headers. If you are storing the data in a mail header, how do you
distinguish a legitimate bounce that doesn't include complete headers
from bogus bounces?
Well, you don't :-) It's a limitation, just as VERP has limitations. You may
lose some bounces from crap MTAs, and only in store-and-forward situations
(i.e. the crap MTA has accepted your mail with 250, then decides to bounce
it later)
I think it unlikely that "most" bounces don't quote headers; surely the most
widely used MTAs (sendmail, exim, postfix, qmail, courier) all do this? And
if you want to "fix" SMTP, then making it a requirement that bounces include
the headers of the original message would be a small change.
Anyway, it's something to play with. Real-world experience and numbers count
for a lot more than theoretical speculation in my book.
Regards,
Brian.