spf-discuss
[Top] [All Lists]

Re: [IETF] Allocation of the new RR type for SPF

2004-11-11 10:10:46
In <20041111141538(_dot_)GA712(_at_)laperouse(_dot_)internatif(_dot_)org> 
Stephane Bortzmeyer <bortzmeyer(_at_)nic(_dot_)fr> writes:

Yesterday, I had the pleasure to present the DNS issues of SPF to the
IETF DNS Extensions Working Group (also known as "namedroppers").

Thanks for doing this for SPF!

There were two problems raised:

* one is old: the coexistence of the two RR types and the transition
period. As all SPFers and MARIDers know, this question has been hashed
and rehashed so many times that most people no longer want to hear
about it :-) Nevertheless, [snip]

I don't want to hear about it.   ;-)


* one is more recent: the draft currenlty defines the new RR type by
saying "identical to TXT format". Some peope (like Mark Andrews, ISC,
one of the authors of BIND), felt that TXT format definition is not
clear enough (specially with the catenation of two strings). Most DNS
RR types have a clear structure (like the MX type, which has priority
and server name). SPF would be the only one with free form. [Because
SPF is a mini-language, it seems the only possible approach to me.]

libspf2 byte-compiles SPF records into a compact binary form.  This
form gives a clear structure and makes SPF records use fewer bytes on
the wire.  So, it *is* possible.  For example, hotmail.com currently
uses 922 bytes of SPF records spread over 5 individual records.  The
libspf2 format uses only 258 bytes and could fit in a single SPF record.

However, there has been very little interested expressed in this and
and I doubt that it would be widely adopted.  I really doubt that the
bind folks would be willing to put a complete SPF parser into their
name server.  Also, because many languages don't deal well with binary
structures (e.g. perl, python, etc.), SPF implementations in those
languages would actually be slightly more complicated.


-wayne