OSEC

Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com
 
Subject: Re: Redhat Linux, Mailman and Sendmail
From: Bennett Todd (betrahul.net)
Date: Tue Apr 04 2000 - 13:23:32 CDT

2000-04-04-14:05:02 Jonathan Bartlett:
> Ummm, RFC 1123 is kind of a long document, care to be more
> specific?

I believe he was referring to RFC 1123 section 5.3.3 "Reliable Mail
Receipt", which begins:

         When the receiver-SMTP accepts a piece of mail
         (by sending a "250 OK" message in response
         to DATA), it is accepting responsibility for
         delivering or relaying the message. It must take
         this responsibility seriously, i.e., it MUST NOT
         lose the message for frivolous reasons, e.g.,
         because the host later crashes or because of a
         predictable resource shortage.

He claims to believe that any non-zero possibility of losing a
message due to a system crash is strictly prohibited by this RFC.

He ignores the possibility (fact, in some of our experience) that
the likelihood of losing data on an ext2fs with default async writes
is way lower than the likelihood of losing the entire drive. Or
perhaps since having a drive toast itself isn't expressly forbidden
by that RFC, he doesn't worry about that even if it's far commoner
than losing data due to system crashes with ext2fs asynch updates.

Whatever, it's not a fight worth pursuing. I really recommend
dropping it. Brad knows a bunch of useful stuff, read that and
ignore the rest.

For whatever it's worth, Weitse seems to agree with Brad on this
one. But fortunately he puts the performance-crippling frob in a
shell script where it's easy enough to dike out:-).

-Bennett

  • application/pgp-signature attachment: stored