Re: [dnsext] [spfbis] Obsoleting SPF RRTYPE

[Doug Barton <[e-mail]>]

On 04/25/2013 03:41 PM, Pete Resnick wrote:
> and I'd really ask you to read through the discussion of issue #9 on the
> list

Ok, so I've read through those threads, and the antecedents. I saw the 
following:

1. Some reports of people who wrote code to use the new RRtype early on, 
and had problems, so they stopped.

2. A report that Perl's "Mail::SPF (which is used by Spamassassin) 
checks for Type SPF first by default."

3. Several folks pointing out the valid DNS protocol based reasons why 
overloading TXT is bad, and new RRtypes for new features are good.

4. Lots of messages along the lines of, "We've always used TXT for this, 
so we should keep using it." These arguments all seem focused around the 
concept that having 2 ways to do the same thing is kind of a pain, so 
let's just use the one that is most popular.

5. Some fairly persuasive technical arguments from Andrew Sullivan that 
putting SPF records into your zone is a good idea. Particularly this:
https://www.ietf.org/mail-archive/web/spfbis/current/msg00544.html

#1 was an expected result in the days prior to 3597 (2003), but should 
be little to no trouble now.

#2 Speaks in favor of my proposal.

#3 (Arguably one of the only "reasoned technical arguments" in the 
bunch) seems to have been ignored by the majority of list members.

#4 Is not a "reasoned technical argument."

#5 Is also a "reasoned technical argument," which was not only ignored, 
but twisted on its ear by at least one list member.

It's also worth pointing out that a non-zero number of list members were 
ready (eager?) to shut down the SPF record prior to their being any 
actual discussion of it.

So I stand by my original proposal. We should do the right thing here, 
not the expedient thing. And the fact that a significant piece of 
software is already doing that, and clearly it works, is a pretty big 
point in favor.

Further, there were no "reasoned technical arguments" in that thread 
against doing what I proposed. As I mentioned in #4 above there was some 
whinging about having 2 ways to do the lookup, but as I pointed out 
previously if the switch had been flipped to do the SPF lookup first at 
some reasonable time post-3597, we'd be celebrating the deprecation of 
the TXT record right about now.

And further than _that_, Andrew and a few other list members pointed out 
that with a theoretical v=spf3 on the horizon, that would be a perfect 
time to say "Use the SPF record only, don't use TXT at all." Of course 
that whole line of thought was shot down with the same, "But we've 
always used TXT" argument. Which is unfortunate because I think that 
would have been an excellent compromise position, which has valid 
technical grounds.

Doug