Per-Entity Metadata Working Group

["Cantor, Scott" <>]

> If DNS has most of the characteristics of what we want, why aren't we
> starting from "act like DNS but for different payloads"?

Partly because the world now is HTTP only, and partly because there's a lot 
of code and specs already written. Also partly because I think the DNS model 
is what it is because it's hierarchical, and that effectively means we would 
just bypass the DNS model altogether and go straight to endpoints hosting 
metadata. I don't think we'd stop at the interim step of every network 
boundary having to operate this service since that's a non-starter.

> The attempt to stand the architecture on its head is, IMO, a useful exercise
> to see if it makes sense. I just keep coming back to "this looks like a 
> problem
> DNS solved."

I am starting to think in fact it has not, when you consider the complexity 
of DNS caching by servers and clients. This is something I'm poking at now 
because we're moving to a GSLB situation at OSU right now.

-- Scott