Re: [syndication] OCS Version 0.5 draft

["Bill Kearney" <wkearney99@hotmail.com>]

> Reptile supports OCS but as we see more-and-more adoption of RSS we are going
to
> see the downloading of whole OCS feeds as overkill.  Right now syndic8 has 11k
> feeds and I just see this getting bigger and bigger.

But there's no need to download the feeds in bulk.  It's possible for user to
give syndic8 a list of feeds and that can be downloaded by others in OCS, RSS or
opml.  So if you're going to have a network of nodes that want to exchange, have
them exchange what's a interest, not just what's known.

> Instead of using syndic8 (or another provider) to fetch a 400G OCS file we
would
> use our 'subscriptions' to discovery other RSS feeds.

Again, Syndic8 doesn't want to be doing that.  We don't want to be pushing the
entire list to everyone.  The expectation is that people can share a list and
others will consume that subset, perhaps building and offering a subset of their
own.

> This scales very well.  RSS weblog communities can be used to keep the signal
to
> noise ratio very high.  The chance that a necessary feed is going to be N
(where
> N < 4) degrees from the current node is very good.

Weblog signal to noise ratio could hardly be considered high...  But the point
vis a vis the similar reader interests, is likely to be true.  So have them
exchange OCS fragments in parallel to a feed.  Both the feed and the OCS may
have different update frequencies.  The OCS file might be changing frequently
but the content hasn't updated in days or vice versa.

> If this is the intent of OCS than I am +1.  If not then I still want to push
> mod_subscription forward as I really want to keep the RSS channel discovery
> mechanism function in a P2P manner..

And you can with OCS, without cramming everything into an RSS feed.  How about
using your mod_link to indicate an available OCS representation of channels
"most interesting" to the RSS feed?  This way a reader program doesn't have to
pull all the subscription cruft when it wants to read the titles.  But if the
user is interested, the presense of the OCS link can be presented and retrieved
if desired.  Nothing's stopping the user from defaulting that behavior for
any/all OCS links detected but the reader who /don't/ want that extra bulk are
completely free to ignore it.

-Bill Kearney