Round robin on CNAME

[Nate Campi]

On Sun, Apr 07, 2002 at 07:13:19PM +0100, Jim Reid wrote:
> >>>>> "Nate" == Nate Campi <nate at campin.net> writes:
> 
> If you say so. IIRC your original message spoke about rewriting
> queries and chasing down delegations.

...and how I wanted to delegate via NS records, not a CNAME. This may
have been in the second message, in order to clarify what I was saying.
Either way, I said it.

> Even if there was a discernable speed-up from your scenario, let's not
> forget the maintenance problems that it leads to. How many times have
> we seen people on this list asking why the DNS continues to give the
> old address of their web server because they forgot to update glue in
> the parent zone? This is or should be a FAQ. 

We manage countless other tiny details regarding UNIX, networking 
equipment, load balancers and god knows how many other things in order 
to keep our web services running. We can keep glue current. We do it in
our DNS already.

> And the logical inference
> from your argument is that every TLD should contain the A records for
> every www.delegation.TLD. [Hey, let's put all these names in the root
> zone so that'll make the lookups even faster....] Didn't a single,
> centralised hosts file prove unworkable back in the days of ARPAnet?

This is NOT like the gTLDs delegating www.lycos.com, but the lycos 
nameservers delegating www.lycos.com. You're sliding down a slippery
slope here, I've implied no such thing.

Jim, obviously we should contract your company to run our DNS. Oh, no
wait, your company allows random shmucks to call up and register their
nameservers as popular domain names used by my company. Nope, we'll run
it I think. You guys have cost us enough money.
-- 
Radioactive cats have 18 half-lives.