[Ietf-not43] extensibility

[Eric A. Hall]

on 7/17/2003 8:07 AM Edward Lewis wrote:

> PS - I haven't had time to really think though how this plays for 
> Routing registries.  If anyone can contribute thoughts on this, I'd 
> appreciate it.

I thought up a good analogy this morning -- the use of multiple IRR
servers with [crisp] is very similar to the use of blacklists in DNS,
where each blacklist essentially acts as an application-specific private
partition, with the clients of those private applications using those
private partitions (where 'using' means knowing about their existence and
sending queries to those servers, versus using the public delegation
hierarchy). IRR servers (and other application-specific uses) would
probably operate in the same basic way in [crisp], with clients being
aware of the partition they need to use (the dc= hierarchy, in LDAP) and
also having to know which data-types to request, how the query should be
formed, and so forth.

We should also expect (and hope) that there would be lots more variations
on this theme, since we would not be limited to [name|type|class] queries
like DNS. Some applications will want to create their own partitions, but
others could reuse the public namespace and simply layer on their own
objectclass definitions at the public leaf nodes, too.

Hopefully that's clearer.

-- 
Eric A. Hall                                        http://www.ehsco.com/
Internet Core Protocols          http://www.oreilly.com/catalog/coreprot/