Re: [xml-dev] RSS beyond the Blog: 1992 or 1999? - was Re: [xml-dev]hur
Lists Home |
Date Index |
Joshua Allen wrote:
you WOULD use RSS syndication for record level notifications.
So you don't see a scale issue with that?
Well, I think RSS is best used for situations where you want to keep
track of events for a relatively broad group of participants which are
updated relatively infrequently. The classic example is monitoring a
list of 20 friends who update their individual websites once every two
weeks -- aggregators help, because you don't want to spend hours a day
trolling through all of those web sites manually, especially when the
majority of the page hits show no changes.
Joshua has characterised it nicely - for a certain set of applications,
RSS works extremely well. For other applications (like the roles of
regular person-person email or IM) it's not so good.
But I believe the range of possible applications hasn't yet been
explored at all. Sure, there is potential for RSS being used as-is
things like bank statements, but the kind of payload currently being
carried is primarily intended for a human audience. Where I think it
will get really interesting is where the payload/data/metadata is more
focussed towards machine interpretation. Take the bug fix announcement
use case. If you get notification that a significant problem has been
fixed in a certain application, then you probably want to patch your
local installation. Why not let the machine do the patching
I think the danger with the syndication technologies is letting the
immediate shiny benefits of newsreader-like apps blind us to the
potential the system has across the board. The RSS 1.0 as RDF
initiative didn't really catch on (too early), but now there is a
teeming demand for extensions to RSS, and RSS 2.0 isn't really up to
the job, being a classic stovepipe application. But it is, along with
OPML, being used in various places as a kind of toy Semantic Web. IMHO
this is a cul-de-sac best avoided. The RSS water has been muddied by
the politics, time has passed, technology has moved and now we need to
move on, and at the same time open our outlooks a little.
If Atom gets solid extension mechanisms together with a normative RDF
mapping (both of which are on the cards) then there will be no seam
between syndication and true Semantic Web applications. RDF/OWL
technologies can offer relatively sophisticated modelling and
processing facilities to syndication; Atom can offer neat RESTful
publication and subscription mechanisms to the Semantic Web.
An additional benefit of Semantic Web technology interop for
syndication is that it provides a framework for message filtering and
hence routing. If you can effectively filter feeds and individual posts
down on topic, person or any other axis then you have at least part of
the solution to scaling. Given that the filtering can be done
dynamically based on user requests, there can be much more efficient
shifting of bits, driven by what's wanted rather than by what's