On Wed, Feb 07, 2001 at 08:56:16AM -0000, D. J. Bernstein wrote:
> I recommend that the A6 and DNAME proposals be terminated. I've set up a
> web page on this topic:
>
> http://cr.yp.to/djbdns/killa6.html
Yes, read, and understood.
How is the example you show different from what is warned against in
RFC 2874, paragraph 5.1.2 ?
OTOH, if you want to eliminate protocol-(cache-) level indirection
altogether, (including CNAMEs (and NS) records pointing to names),
you'd need to write a draft of all of it up. (This includes an
address-valued NS replacement.)
And by providing a formal description of how the servers resolve the
indirections, and how renaming has to happen if this is implemented, we
will be able to check whether this will work, and what requirements are
needed to make this work (e.g., how to ensure convergence when renaming
or renumbering happens for one of the components). I'm not yet convinced
that moving the indirection from the caches to the servers helps in all
cases.
Regards,
-is
PGP signature
