Re: APL and other new RR's

[Kevin Darcy <kcd@daimlerchrysler.com>]

I realize it's a bit abusive, but why not use SRV records for this? The
preference fields could be used as an index by which to associate the
addresses with their masks. Heck, maybe the preference field could _be_ the
prefix length, thus eliminating the need for separate address and mask
subdomains.

IMO, the root problem here is that DNS has no order-dependability, period.
Perhaps DNSEXT could address this general problem, come up with a general
solution, and this might obviate the need for all of these specialized
solutions.


- Kevin

Paul A Vixie wrote:

> DNSEXT and its predecessors seem to deal just fine with new RR's which have
> to do with the DNS itself (as in SIG, KEY, OPT, and TSIG come to mind here.)
>
> DNSEXT as an "oversight committee" for new RR types which originate elsewhere
> seems to work fine (NAPTR, SRV, LOC, NSAP, NSAP_PTR, are all examples) seems
> to work.  All DNSEXT is being asked to do in those situations is ensure that
> the RDATA encoding and other DNS-relevant issues are right -- NOT determine
> whether the proposed RR is going to be useful or whether the design that it is
> part of is really the right way to build some Internet application.
>
> But trying to get something like APL "started" inside DNSEXT is not working,
> because the implicit question is "what has this got to do with the DNS?" and
> the implicit answer is "nothing."
>
> I have an application that needs distributed, reliable, coherent, autonomous
> access to a list of CIDR blocks.  I would like to store these in the DNS since
> the DNS is already a distributed, reliable, coherent, autonomous "database."
> My application has nothing to do with the DNS protocol.  DNS's existing types
> don't deal well with this.  A RR's don't have a mask.  I can't use subdomains
> like ._addr and ._mask because then a single RRset won't have the whole list
> (._addr and ._mask would be separate RRsets and are not order-dependable.)
>
> Apparently, from comments posted here in response to the APL RR draft, most of
> the DNSEXT WG does not understand why this is being discussed in DNSEXT at all
> since it has nothing to do with the DNS protocol, any more than NAPTR or NSAP
> does.
>
> Therefore I hope the author will withdraw the draft and reoffer it as an
> individual, non-WG product, on an "experimental" track, which is what worked
> when SRV ran afoul of this same lack of support.  (I'm just sooooo sure that
> Microsoft worries about whether Win2000's SRV support is really
> "experimental".)





to unsubscribe send a message to namedroppers-request@ops.ietf.org with
the word 'unsubscribe' in a single line as the message text body.