Re: practical matters (was Re: tired of waiting)

[Erik Ostrom <eostrom@research.att.com>]

> on the namespace issue, yep, this is a tricky one.  one reasonable
> possibility is to simply name a namespace czar, and let them handle
> registration.  vijay made a good suggestion to me about allowing
> people to name stuff by prepending some unique name, like a DNS name.

Might be better to go with DNS name and some unique ID (such as a user
id).  I think I would hate it if I had to share the com-att-research
namespace with all of AT&T labs, and picking a random machine here
doesn't really help.  com-att-research-eostrom admittedly seems
egocentric; but com-att-research-twin for the twin protocol seems
appropriate.

in the interests of brevity, i'd vaguely prefer a combination of the
DNS scheme and the registration scheme--anyone can experiment by using
DNS and a unique id, but parties can contact a registrar to get a
piece of core namespace.  (maybe, only if they provide a published
spec?)

> in any case, i think i'd like to reserve the protocol namespace under
> 'mcp-' to the protocol.  is this reasonable?  if so, i'll add that
> too.

please.