Re: [syndication] My take on shared feed lists

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

> No doubt I'll regret stepping into this fracas, but I'm trying to wrap
> my brain around the protocol/format. Since we fetch over 31k feeds every
> hour, and will want to support whatever comes of this, maybe walking
> through what we'd have to do in support of this would be educational. At
> least to me. ;-) I'm guessing that other aggregators will do roughly the
> same as what we'd do.

Excellent feedback from actual users and implementors is always good.

> The subscription file won't change very often, so we'd only try to get
> it every N days (or maybe only once). What would be involved in us
> fetching that? Well, for each site, we have a normal url and a feed url,
> as specified in the RSS file. Our options are to fetch the HTML from the
> normal url and parse that looking for an appropriate <link> tag, or to
> try to directly fetch the subscription file, using the feed url and a
> standard naming convention.
>
> Obviously, in this case, much less bandwidth will be used if there's
> just a standard naming convention and we don't try to look for <link>
> tags. We'd only have to fetch one file.

Unless it's not clear what 'level' of the URL pathname at which it's to be
found.  Then you'd end up on fishing expeditions, blindly stabbing for files and
possibly getting the wrong ones.

> I'm not saying that's necessarily the best solution. A standard naming
> convention of course has its drawbacks, and all in all I'm undecided
> whether I like it or not. But if we are to discard or deprecate a
> standard naming convention, I would *strongly* suggest augmenting the
> various RSS protocols so that a feed can specify the location of this
> file. That way, we still get the benefits of only having to fetch one file.

Now that's a fine idea.  Put an element in the RSS that also says where more can
be found.  In RSS-1.0 this is very easily done.  In 0.9x it's not supported.
2.0 is anyone's guess.  The use of modules benefits greatly from clarity on the
use of namespaces.  That's still not been resolved effectively in the pseudo
spec.

-Bill Kearney