Re: [dnsext] Obsoleting SPF RRTYPE

["John R Levine" <[e-mail]>]

>>> Once again, the huge practical barriers to deploying new RRTYPEs made the SPF RR dead on arrival.
>
> John, I completely disagree, as you know, with this statement. What 
> killed SPF RR was the introduction of the TXT kludge for SPF.

You're confusing cause and effect.  For one thing, back in 2004-2005 
getting a new RR type allocated was so painful that it wasn't worth the 
effort when they were trying out SPF.  I realize this problem has been 
somewhat alleviated.

The much more serious reason is that the provisioning systems to let 
system managers install type 99 records into their DNS did not exist, and 
for the most part still do not exist.  I know a lot of the people who did 
early SPF installations and they uniformly reported that even getting the 
provisioning software to support TXT records that have been around since 
the dawn of the DNS was painful, and there would have been no hope for a 
new and exotic record type.  Since their goal was to make SPF actually 
work, they used a kludge.  It's the same reason DKIM uses TXT records.

>> Yes, the ossification of the DNS makes introducing new things 
>> challenging however as Mark pointed out, software was beginning to do 
>> the right thing and there actually are web interfaces out there that 
>> let folks enter SPF records (I use one).

Wow, and it's only taken eight fripping years.  Perhaps if we wait another 
century the provisioning systems will catch up.  If the DNS crowd spent 
half the effort addressing the provisioning problem that you do trying to 
deny that it exists, it'd have been fixed years ago.  My dnsextlang 
proposal is one approach, but surely not the only one.

In any event, the SPF draft is in WGLC.  Feel free to read the extensive 
discussion in the list archives and let them know how you feel.

R's,
John

PS: The system that lets you enter SPF records wouldn't be Godaddy, would 
it?  If so, create an SPF record, then do a dig and try to find it.