Caching does not change the underlying semantics. It is merely an
efficiency hack added to that semantics.
The semantics of the setup is a DNS query.
Cache it once, subsequent requests (even weeks later) use TCP/IP
without resorting to DNS queries. The effect for most (individual)
lookups is the same, even (for instance) after DNS is poisoned.
If you start with a domain name, then the first step is the DNS,
from a formal standpoint.
Hmm . . . but *which* DNS system?
(Theoretically, one can have a domain name that is not tied to a DNS
query; that is, a distinction between a registration in the name
space, versus a query event that seeks to map a registered name to
some values. But that doesn't happen in the real world.)
I could sort of see it as a pre-XRI hack to support Tor (which the
mainstream DNS *doesn't*), but that's hardly "the real world", so
never mind ;)
OpenIDs that don't use any domain names would, yes, be an example of
an alternative.
No domain names, or strings that look like domain names but *can't*
be looked up through the usual DNS?
-Shade
_______________________________________________
specs mailing list
[email protected]
http://lists.openid.net/mailman/listinfo/openid-specs
