Re: [syndication] Is a Feed the right place for your Data?

[Ken MacLeod <ken@bitsko.slc.ut.us>]

"Dave Winer" <dave@userland.com> writes:

> This isn't for aggregators, but if I wanted to load my weblog into
> another program, I could just loop over all the days and get the RSS
> data. It's a data interchange format. That was the point of my
> post. Having a format that models a day's of weblog posts is cool,
> and all these formats we discuss can do exactly that. Dave

We've recently been discussing this in the Atom context, but it
applies equally to RSS.

The issue/suggestion originated in a discussion about "search".  For
Atom, we'd already concluded that search results are no different than
a feed, so search results use feed format.  Another issues was
querying for "all entries" -- for some sites that could be thousands
of entries, way too big for one result file.  Yet another issue was
that searches by nature are usually dynamic, when the result set for
now-archived entries could easily be made logically static (which has
HTTP and web benefits, it doesn't mean they can't be dynamically
generated anyway).

While there is still a need for a dynamic search for other purposes,
for "last N entries" or "all entries" it would be far better if we
extended the already existing feed format to include a "previous"
link, a link to the "next" feed backwards.  There's no definition on
how big an archive feed needs to be, it could be a day, a month, 100
entries, whatever.  Aggregegators, thus users, can set their own
thresholds, "when subscribing to a new feed, pre-fetch last 200
entries".

In Atom, the current proposal is for the <feed> element to contain an
X/HTML-derived,

  <feed>
    <link rel="prev" type="application/x.atom+xml"
          href="http://example.org/feeds/2003/October.atom";
          title="October 2003 Atom feed" />
    ...

Again, the link and title would reflect the publisher's preferred
scope for keeping back-linked archives.  Something similar could be
developed for RSS*.

  -- Ken

* There is mod_link, nominally for RSS 1.0, but with my RSS/RDF hat on
  I still think a named property and/or rdf:seeAlso is better than a
  "link language" (mod_link) layered over a what is already a "link
  language" (RDF).
    http://www.peerfear.org/rss/permalink/1027556725.shtml