Re: [syndication] Re: RSS feed filtered by keywords?

["Bill Kearney" <ml_yahoo@ideaspace.net>]

> >FWIW (in agreement), unless something fundamentally changes in the
> >interim, when we have to deal with scaling to 200M or 200G feeds the
> >bottleneck will not be storage or processor power, it will be bandwidth.
> >
> Or really ... what\where is the bandwidth bottelnect that you are
> refering to ?

He's correct.  Feeds are now hovering around 80mb a day of text.  This isn't
much on the whole. But if more people try to read more feeds, the aggregate
effect would be a bandwidth problem.

> Don't get me wrong .. I know of no reason that RSS is not scaling just
> fine ... I just know that it is already beyond the point where I can
> personally read all the items flowing in the RSS river ... in fact I
> can't even keep up with all the items in the feeds to which I have
> personally subscribed.   Now we can rely on  Google , and personal
> browsing for discovery  ... but that is slow and will miss a lot of the
> current flow.   How do we go faster and beyond what Google and personal
> browsing can offer?   Can you  reliable and swiftly give me all the
> items that have talked about identity as it relates to formal logic on
> the semenatic web in the last week ?

Good example, a keyword search for 'semantic web' would miss your posting.
Spelling's a bitch, ain't it?

Fortunately folks are working on this.

What the feeds can do to help is be just a little bit more informative in what
they output into their XML files.  Stuff like contact info (spammers be damned)
helps get a picture of who's reading what.  Using my suggestions on geographic
info helps figure out where it's being said.  Using W3C standard date formats
figures out when.  Applying semantic web principles helps flesh out the bigger
picture.

So as to your example, I'd start with you.  You've got a foaf file, right?  If
so that can be used to build out a larger picture of who else to 'overlap' in
some of the searches.  You share your channel subscriptions, right?  Using that
I can do comparisons of it against mine and the other people (via foaf) that are
involved in these circles.  See how this is going?  I'm nowhere near bothering
with the raw RSS itself yet.  Step back and see if you're offering up an XFML
document.  That'd be a good way for me to see what topics and facets have been
of interest.  Again, no content yet!  Apply the same overlapping to that set of
people that's being built out.  Perhaps even recurse down into comments for even
more linkage.

Now, armed with this overlapping set of factors it becomes a lot more practical
to *know* where to start looking for more info.  The RSS here is undoubtedly
going to be a much smaller subset.  Keyword filtering would likely not be as
necessary at this level.

The same faceted 'pivoting' can be done from any starting point.  The trick is
it depends on people offering up the useful bits of metadata.

-Bill Kearney