Missing the point: Service discovery protocols.

Christopher Granade cgranade at greens.org
Tue Nov 1 16:18:13 PST 2005

Hash: SHA1

In recent threads, there has been much argument as to what the "best"
method of returning metadata about an identity to a consumer. I think
that the point is being missed, as there is being made the assumption
that one method could ever serve all use cases.
Unfortunately, one size does not fit all. My needs are different than
your needs, or anyone else's needs. What we need to do is to identify
different use cases and specify an acceptable meta-data protocol for each.

To start the process, let me enumerate several such cases. Each is based
on the example URLs
(A) http://www.example.com/~jdoe/
(B) http://www.example.com/~jdoe/?meta=capabilities

In order to do this, I will consider each of several properties, and the
set of use cases will be the full set of possible combinations of these

(1): Always applies.
     (a) Page is HTML.
     (b) Page is XHTML.
     (c) Page is other XML type.
     (d) No page exists at (A).
(2): Always applies.
     (a) Server cannot distinguish between (A) and (B).
     (b) Server can distinguish between (A) and (B).
(3): Applies only if (1a), (1b) or (1c) is true.
     (a) Page is static.
     (b) Page is dynamically generated.
(4): Applies only if (1a), (1b) or (1c) is true.
     (a) No other metadata is to be applied to (A).
     (b) Significant amounts of other metadata are to be applied to (A).
(5): Applies only if (4b) is true.
     (a) Other metadata is not in RDF.
     (b) Other metadata is in RDF.
         If (1a) and (1b) are both false, embedding assumed.
(6): Applies only if (5b) is true.
     (a) RDF is included via <link rel="meta" /> element.
     (b) RDF is embedded via <link rel="rdf.type" /> elements,
         representing RDF statements where the subject is implied.
(7): Always applies.
     (a) Performance is more important.
     (b) Ease of maintenance is more important.

For each combination, the consumer should understand a format that is
appropriate for the use case. Note that I assume that the consumer is
more able to make concessions to the identity than visa versa.

It is my assertion that an optimal set of solutions would be:
(i):    If (6a), then add extra metadata to existing document.
(ii):   If (4a) or (6b), then add extra <link rel="" /> elements.
(iii):  If not (1d) and (3b) and (7a), offload metadata to
        static RDF document, and reference via <link rel="meta" />
        element for (X)HTML documents, or via XInclude otherwise.
(iv):   If (1c) and (3a) and (7b), embed metadata directly as
        RDF as XML.
(v):    If (1d), return RDF as XML or RDF as N3.

Sorry for the long post, but I wanted to carefully enumerate what I have
been trying to say here.
- --Chris
Version: GnuPG v1.2.4 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org


More information about the yadis mailing list