SPF on 9.4.1 now?
milli at acmeps.com
Tue May 22 01:12:37 UTC 2007
Mark Andrews wrote:
>>Mark Andrews wrote:
>>> No. You use it *instead* of TXT record. There is no need
>>> to dual publish the data. Anyone that really cares about
>>> SPF will upgrade their clients.
>>As a practical matter, I must respectfully disagree. It will be some
>>time before everyone gets a chance to upgrade, and the timeout issue
>>with looking up SPF from some DNS server sets (not BIND or MS
>>implementations far as I can tell) is a significant issue. This timeout
>>issue could, of course, be a firewall issue... anyway, it has a
>>significant impact on high-volume (for various definitions of "high")
>>mail sites. And thus is ultimately off-topic for this list. FIN.
> What timeout issue? If you don't publish the old clients
> will get a NODATA response. There is no time out issue in
> not publishing the TXT record.
The timeout issue is with looking up SPF records on some name servers.
$ dig +norecurse TXT massivebonus.com @ns1.massivebonus.com
$ dig +norecurse SPF massivebonus.com @ns1.massivebonus.com
to see what I mean (with a 9.4.x version of 'dig' of course). And
nevermind that this example domain is quite borked in its
delegation/glue path. This is an issue with a lot of domains that
source spam and have SPF TXT records to make the envelope sender look
"good". This timeout issue will slow down adoption of native SPF
SPF does not stop spam, the spammers were the early adopters of SPF...
Again, this is all off-topic here. Just pointing out that, like any
other change like this, adoption of _the_ SPF Resource Record instead of
using TXT records is not going to happen overnight because TXT records
with SPF info are out there already, and also because the lookup of SPF
instead of TXT is out of the hands of the organization that cares that
other MTAs look them up... chicken-and-egg or catch-22 apply, and thus
to maximize the chance that _everybody else_ will find SPF information
and act on it, smart administrators will put in both types of records in
the short term. That's what I see happening in my crystal ball, like it
or not. :-/ I'm all for a flag day! It's just unlikely to happen, and
more unlikely to happen because of this timeout issue, IMO.
Michael Milligan -> milli at acmeps.com
More information about the bind-users