On Fri, 28 May 2004, Bob Atkinson wrote:
I not against the use of wildcards; wildcards per se are wonderful so
far as I am concerned. I've never met a wildcard I didn't like; got
nothing against 'em.
Oh, good. Then we have a consensus that wildcards are just so wonderfull !
It's just that, as I understand things, if we are to accommodate them,
then we need to put our MARID data in the domain's entries itself, not
in a subdomain. If I'm wrong about that, please let me know.
We do have a compromise proposal that records go into _marid.domain.com by
default but that those wishing to support wildcards should be allowed
to add them into domain.com if they wish to support _marid.host.domain.com
and other situations. Since we can speculate that wildcards are not the
common behavior, majority of dns records would have TXT in its own
semi-namespace.
I have to point it that even if we dont specificaly say this, doing it
would still be perfectly legal and ok with dns, so I'm certain people who
want willdcards will quickly find that out and do it anyway.
And it's that that I have a problem with. Using TXT records in the raw
domain level
a. conflicts with any existing TXT information that is there, and
I agree. I'll go futher then that and say that using TXT in general conflicts
with existing information that may possibly apply for that domain
b. will interact very poorly or not at all with the next batch of
information that some future standard might wish to publish about a
domain, building on the precedent and education that we will set about
how to go about doing that.
We're actually the ones creating bad example by choosing to use TXT records.
If we're to do it, it should be clearly said that new type will be created
and that every company involved will make their best effort to quickly
build support for this new record type.
So I understand there to be a cost in choosing to support wildcards.
As opposed to what? Benefits that come with using them or costs to
companies that have used wildcards for MX and now have to do it completely
different way for "reverse-mx" records?
I thus am lead to try to understand the actual relative costs and
benefits of doing things one way vs doing things the other. All I'm
trying to do is get real data about this, from actual real world
experience as much as possible.
And this may not be right group for that practical experience.
Too many here are engineers concerned with protocol designs.
PS: You can keep MSFT-bashing if you wish; I'm used to it. But I truly
am trying to be constructive here.
I suspect that in the eyes of dns community the entire MARID WG and all
participants here deserve some serious bashing except possibly if we use
something other then TXT from the beginning (and even then they'd be happy
to bash us for doing level9 system design at the IETF WG level) ... :)
--
William Leibzon
Elan Networks
william(_at_)elan(_dot_)net