RE: URNs, Namespaces and Registries

Hi All,

Thanks to David Booth who was nice enough to spend some time with me on
the phone to help me understand some of the suggestions in his e-mail
[1] and in his urn2http paper [2]. 

In one of my prior messages [3] I agreed that "Domain names are as good,
or as bad, at conveying _ownership_ of a particular form of URI as URN
namespaces or URI schemes", but then I raised questions about other
aspects besides ownership. 

One question was about "Syntax Aspects" of a form of URI, like assigning
special meaning to characters that are unreserved in http. After
speaking with David, I agree that using an approach like
"http://xri.net?<rest_of_XRI>" could let an application know about the
special characters just as well as using "xri://<rest_of_XRI>". So I
withdraw that question.

Another question was about "Authority Aspects" of a form of URI. My
concern is that with an approach like "http://xri.net?<rest_of_XRI>" the
authority for the identifier likely appears in the <rest_of_XRI> instead
of in the authority segment of the URI. My gut feel is that the
authority for the XRI should appear in the URI authority segment;
however, I'm having so much trouble describing my concern that I'm
beginning to think my concern is not valid. So I'll probably withdraw
that question too.

Another question was about "Operations Aspects". I would still like more
discussion on this question. Not all schemes support the same
operations, and indeed they don't all support HTTP GET. E.g., ldap: does
LDAP bind and search operations, ftp: does some ftp-specific
operation(s?), xri: supports an operation we call XRI RESOLVE, and other
schemes probably support yet other operations. By using specialized DNS
domains instead of separate schemes, do we lose the ability to support
operations beyond those supported by http:?  Or, is it expected that an
application aware of the meaning of a specialized DNS domain like
"http://xri.net" will therefore know about XRI supported operations
(just like an application would have to be aware of the meaning of
"xri://" in order to know about the supported operations)?

As you can probably tell, I'm not as opposed as I used to be about using
http: for XRIs. In earlier discussions and examples I kept hearing that
there is no need for XRI because everything can be done with http, but
none of the examples were satisfying to me (or others on the OASIS XRI
TC). Now I'm hearing that one way to do what XRI does in http is to just
use something like "http://xri.net" instead of "xri://" and keep all the
functionality specified for "<rest_of_XRI>". Such an approach is much
less distasteful to me than saying there is no need for XRI because
everything can be done in http. Can TAG members please clarify if their
gripes about XRI would dissolve if XRIs begin with "http://xri.net"
instead of "xri://"?

Thanks!


[1] http://lists.w3.org/Archives/Public/www-tag/2006Aug/0037.html
[2] http://dbooth.org/2006/urn2http/
[3] http://lists.w3.org/Archives/Public/www-tag/2006Aug/0024.html


Marty.Schleiff

Received on Friday, 11 August 2006 05:42:47 UTC