A records point to a domain

[stanislav shalunov]

Barry Margolin <barmar at genuity.net> writes:

> If step 3 fails because the nameservers don't respond, the MTA
> should give up immediately and put the message back in the queue to
> be retried later, precisely to prevent the above bug.  Why would
> they even consider going on to step 4?

Of course.  I have checked sendmail 8.9 source, and naturally it does
it this way.  However, I could imagine an MTA author who creates a
function that returns a list of MX records or a failure + sets an
error-type global (a la errno).  They could just go on to see if an A
record is available if that function returns *a* failure.

Obviously, this behavior isn't desirable, but it would be hard to
detect it with testing.  So I assumed that there might be MTAs out
there that do it, since it seems a very natural mistake to do, and can
live in the code for a long time unnoticed.

As you point out later, the standards don't explicitly prohibit such
behavior (unfortunately).

-- 
stanislav shalunov				| Speaking only for myself.