At 20:06 13/12/2007, Hector Santos wrote:
Arnt Gulbrandsen wrote:
So if I understand you correctly, you meant to
say: «All current/modern DNS clients will
support the resolution of a CNAME as part of a
MX expansion process, when "current and modern"
is defined to take this issue into account.»
That's just a tautology wrapped in some suggestive wording.
(Gawd, this is so silly.)
Arnt,
I guess that depends on how you view or define "Current and Modern."
I change my statement:
"Industry ready DNS clients support the resolution of a
CNAME as part of a MX expansion process."
Another else, IMO, is mal-practice. <g>
Can we move on?
Hector, the reason I'm 'debating' here is not
because our software doesn't handle CNAME entries
as MX records (it does), but because saying
things like 'all modern' or 'all industry ready'
or anything else which sounds like 'all good' is
implying that anything that doesn't is poor
quality or is breaking standards - when it most certainly isn't.
The debate comes up frequently of whether
software should try to work around buggy
implementations or fail them. Some people think
you should do anything possible to make it work,
others think you should fail anything which is
slightly wrong, and most people are somewhere in
between. One thing most agree on is that we
should TRY to make sure everything works
according to standards even if most software can
work around the non-conformities. That means we
should strongly discourage people putting CNAMES
in as MX records - including telling them they're
wrong if they do. It might work if you do, but
that doesn't mean that it's an OK thing to do.
Saying that "anything which treats a buggy
configuration as buggy is 'mal-practice'" is very
dangerous IMHO. It could lead to people writing
buggy implementations and saying 'xyz email
client handles it so it must be right - ignore
what the standards say, they're not what matters,
what matters is common practice'.
IMHO, 2821bis should state that "putting CNAMES
as the target of MX records is not allowed, as stated in RFC 2181"