siteData-36 and Atom

I've suggested that Atom [1] could use a Site: HTTP header to provide site
info. It's been suggested that this might in some way conflict with the
proposed TAG approach, so I thought I'd better check.

One of the requirements of the Atom system is to be able to easily discover
site services, services that fulfil functions usually found on weblogs :
e.g. posting blog entries and comments, the feed for syndication. The
preferred form of these services is standard http methods. A way of
describing the services has been pretty well worked out in the Atom API [2],
based on the common Atom format. An example would be:

<link rel="service.post"
    type="application/x.atom+xml"
    href="URI for Posting goes here"
    title="The name of the site.">

I've suggested on the mailing list that the services could be advertised
using the approach (described at [3]) of using an additional Site: {uri}
HTTP header to identify the (conceptual) website. When a GET is done on that
URI, an Atom-format file would be offered with the mime type
="application/x.atom+xml" (or similar).

Given that the mime type isn't one likely to be used for other descriptions
(RDF, RDDL), and that the meaning of "Site:" is the same (a set of resources
that form part of what's commonly known as a web site), I believe this to be
consistent with the TAG proposal.

I do have doubts of my own, but I think they are covered by the idea of Web
Site and practical limitations:

* the Atom file wouldn't contain an exhaustive list of all resources on site
* the Atom file may contain additional information/resources

An exhaustive list probably wouldn't be desirable in practice. Additional
information shouldn't have side effects, so for a generic "Site Manifest
Reader" anything else would appear as comments. This latter case may be
significant with Atom, because it does point to the possibility of the Atom
file with the "Site:" URI being the syndication feed (e.g. URIs of the 10
most recent blog posts) as well as containing relatively static site
metadata.

fyi, an (X)HTML <link> element on the site home page with the URI of the
Atom data has already found its way into the (provisional) Atom specs. It's
been pointed out that if the home page is lousy markup, it's difficult to
obtain the URI. Although in most cases an Atom capable system would include
generation of the home page, it's still conceivable that it might be
unparsable (e.g. an end user template messess things up). Hence interest in
alternatives.

So, does this all sound consistent with current proposals relating to
siteData-36?

Cheers,
Danny.

[1] http://www.intertwingly.net/wiki/pie/FrontPage
[2] http://bitworking.org/projects/atom/
[3] http://lists.w3.org/Archives/Public/www-tag/2003Feb/0297.html

----

http://dannyayers.com

Received on Thursday, 8 January 2004 07:25:23 UTC