In
<16566(_dot_)38305(_dot_)82240(_dot_)255682(_at_)cse(_dot_)unsw(_dot_)edu(_dot_)au>
Neil Brown <neil(_at_)brown(_dot_)name> writes:
As I understand it, there are three approaches that have been
discussed.
2/ TXT record in the domain (used by SPF). This is (arguably) the
simplest but collides badly with other uses of TXT.
I disagree that, in the case of SPF, this collides "badly" with other
uses of TXT.
As I published in previous posts about existing usages of TXT and SPF
records, I found:
1) SPF records are small. 50% of all SPF records are under 35 bytes,
99.8% are under 200 bytes.
2) SPF avoids collisions via well defined magic numbers. In my survey
of 1.3 million domain names I could find no other existing uses of
TXT records that seriously conflict with SPF's usage.
3) There isn't a significant install base of TXT record usage. SPF
records already account for nearly 20% of all TXT records at the
domain level in just 6 months.
I would hope that in the future, the adoption of new RRs would be made
much easier by things like updates to MicroSoft's DNS system. I would
also hope that most future needs will not be so pressing as MARID.
Still, if future needs for TXT records are as "well behaved" as SPF,
we can have another half-dozen to a dozen such RFCs before most domain
owners would have to start picking and choosing about which TXT
records they should use.
It is comforting to know that the research, surveys and analysis that
the SPF group did last fall have been confirmed by more recent
analysis.
In short, I think the problems with SPF usage of TXT records is far
more theoretical (and esthetic) than practical.
-wayne