Re: [spfbis] The RRTYPE topic

[Stuart D Gathman <[e-mail]>]

On 05/30/2013 02:43 AM, Mark Andrews expounded in part:
> In message <[e-mail]>
> , "Murray S. Kucherawy" writes:
>> On Wed, May 29, 2013 at 5:39 PM, Mark Andrews <[e-mail]> wrote:
>>
>>> As for numbers the survey was taken *very* early in a transition
>>> from TXT to SPF.  SPF record have doubled as a percentage since the
>>> initial survey was taken.
>>>
>>>
>> RFC4408 was published in 2006, and SPF had been in extra-IETF development
>> for a while before that.  RFC6686 was published less than a year ago.  I
>> don't see how you can characterize that as "early".
> In a project that I would expect to take 15+ years years to complete
> yes it was early.  For a project where the SPF propronent took until
> 2008 to issue new library code that made type 99 lookups.  Nameserver
> vendors incorporated SPF support faster than the SPF propronent.
> Then it required MTA vendors/package maintainers to use the new
> code.  There wasn't even one hardware replacement cycle allowed
> for.  It takes OS vendors years to integrate new nameserver code
> into their releases.
>
> TXT to SPF transition was never going to be quick.  It has/had a
> timescale similar to A to MX only domain adoption.  As far as I
> could see the transition was roughly where I exected it to be.  If
> the SPF working group had wanted a faster transition they should
> have set development milestones and requested that nameservers
> vendors log missing type 99 records etc.
>
> Changing things in the DNS takes a lot of time.  This is something
> DNS developers know and accept.
+1

4408bis makes a provision to endorse caching negative and error results 
for up to 24 hours.  *That* (not throwing away the SPF record) is the 
solution to the broken nameserver problem.  When you query SPF, and the 
request times out due to broken firewalls, broken nameserver, or 
whatever, then cache the result.

If anything, it is time to remove the dual publishing (both TXT and 
SPF), and the administration/consistency problems that causes.