spf-discuss
[Top] [All Lists]

Re: SRS

2004-02-17 08:41:04
In <20040217152657(_dot_)GB24189(_at_)csi(_dot_)hu> 
mw-list-spf-discuss(_at_)csi(_dot_)hu writes:

On Tue, Feb 10, 2004 at 05:02:04PM -0500, George Schlossnagle wrote:
How does this reconcile the reverse-path size limitations imposed by 
rfc2821?

The ezmlm list manager has been breaking this limit for at least 7
years.  Under qmail, any user can easily create addresses with "too
long" localpart.

By "breaking this limit", do you mean that ezmlm *always* creates
local parts with longer than 64 bytes?  Or, does it just sometimes
create them, depending on the particular user involved?  I am pretty
sure it is the latter case.


SRS has a couple of differences between mailing lists that do VERP,
even though they both create long local parts.

First, with a mailing list, any given user will receive a pretty
constant length local part, based on the address they signed up to the
mailing list with.  If this local part causes things to break, chances
are, it will break right away, and the user will either take
corrective action on their end, or give up on the "broken mailing
list".

SRS, on the other hand, will create a many different local parts for
any given user.  A mailing list's VERP encodes the destination email
address, while SRS encodes the source email address.


Second, the SRS system has to not only encode an email address, but it
in the case of SRS1, it also has to encode another domain name.  This
makes it longer than most VERPs.


I'd be surprised if there were any smtp servers out there enforcing
the 64 character limit.

Be prepared to be surprised then.  This has already been discussed a
couple of times, and just last December, there was a case where
someone's long VERP was breaking CISCO's "mailguard" system.


But why not focus on the really fundamental problems:  

I agree that these issues need to be addressed also.


-wayne


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