I think using DNS for datatransfer is an interesting idea,
but I don't think it's a good idea here (or maybe I just
don't get it...)
Maybe I'm missing something, but I see 2 possible uses
for DNS:
1. server to server replication
To my knowledge DNS do not do replication per se. Of
course there are zone-transfers from primary to secondary
DNS, but think about it for a second...
I saw proposals for about a dozen domains (key-to-artist,
key-to-album, toc-to-key,....) Each of these subdomains
(will) contain many thousands of records (cddb has an
index of 30k CD?)
Every zone-transfer sends _all_ entries of a zone...
Another thing I don't like is the fact that only one
primary DNS can incorporate changes. This means one
point of failure, a single point of "power"...
2. query from client to server
This could actually work. Lots of existing code and
infrastructure can be reused.
One argument I heard of is cacheing. That would be nice,
but http is cacheable too and I don't think there's much
to cache anyway (do you listen to the same cd all the time?
I think not. The chances to reuse entries before they're
expired in the cache are quite small...
Another negative point is the load on the dns-system. It's
one thing to create lots of http-traffic, but it's another
thing to (ab)use nameservers for something new which they
didn't expect to do. I'm not shure they'd like the additional
load...
But the killer-argument are firewalls. Many firewalls do
block nameresolution (yes, I know you can surf the web and
nameresolution seems to work for you, but it's actually
the web-proxy doing the dns-lookup, not the client).
So my conclusion is to drop the idea of using DNS for our
project here.
Bye
dworz