Some years ago I saw a long essay posted on the web with a
like "DNS makes a bad distributed database for things other
names". It was written in the style of an RFC, but it doesn't
look like it
was ever made part of the series.
I seem to recall John Klensen getting irate about all sorts of
junk just getting thrown in without any sort of architecture.
There was a presentation at an IETF.
I think that a lot of his points were on the issue of fuzzy match
though. DNS is fast because it is an exact match scheme. Try to
fuzz it or include search (aka X.500, LDAP or other failures)
and it goes bazzzongo.
I agree in principle, but disagree that there is no architecture.
We could have one, just insist that the SPF TXT record uses
the _SPF prefix.
It is an entirely logical extension to the DNS. Instead of
extending through the RR system we extend through the name
I would add that the use of DNS for distributing RBL information has a
remarkable breadth of support covering all major open source and
proprietary MTAs, and could better be held up as a model for
rather than carping about minor shortcomings.
The other information point here is that there has been
a lot of work on Web Services and WSDL and WS-Policy and
that can all be run back to the DNS through appropriate
Asrg mailing list