Good stuff Julien ! My 2c: When I was working at MySpace we designed an in-depth filtering mechanism. You can see it ta : http://wiki.developer.myspace.com/index.php?title=Stream_Subscription_Query_Spec external to the PubSubHubbub spec. We spent significant resources iterating with developers on whether their permutations of filters were valid and never got around to providing a test tool. Facebook started with a far simpler subscription model that only supports specifying 2 fields to filter on (http:// developers.facebook.com/docs/api/realtime) but also has no test tool.
After seeing developers wasting time figuring out how to filter/ subscribe to things my personal recommendation is #1 To use the topic url so its testable with a browser. #2 Which can be done outside of the spec, for hubs to provide "Virtual" Topic Urls. One example of this would be a firehose url with parameters for filtering. Also outside the scope of the spec would be for hubs to promote their virtual topic urls like Julien just did. This will be helpful as subscribers don't have to subscribe to raw topic urls one by one and can use the virtual or aggregated topic urls from the Service Provider Thoughts ? On Jul 1, 10:13 am, Julien Genestoux <[email protected]> wrote: > Hey, I don't mean to interrupt... but maybe this is part of the conversation > : Superfeedr announced its track feature today > :http://blog.superfeedr.com/track/filter/xmpp/pubsubhubbub/track/ > > You can subscribe to keywords accross all the feeds that we host (more than > 800 hubs now) or the feeds we have for the default hub. > > We want to get more filtering done... but it's obviously a quite complex > task with high frequency publishing! > > Cheers, > > julien > > On Thu, Jul 1, 2010 at 7:04 PM, John Panzer <[email protected]> wrote: > > A topic: URI scheme? > > -- > > John Panzer / Google > > [email protected] / abstractioneer.org <http://www.abstractioneer.org/> / > > @jpanzer > > > On Thu, Jul 1, 2010 at 5:22 AM, Alexis Richardson > > <[email protected]>wrote: > > >> On Wed, Jun 30, 2010 at 10:30 PM, Danny Briere <[email protected]> > >> wrote: > >> > This is interesting, because I think we may be more concerned with the > >> > "topic model" than the content filtering. So we have some questions > >> > here: > > >> > 1) What is the current thinking / discussion / issues about the topic > >> > model? I've searched the recent postings trying to find references > >> > and more specifics....is there a URL I can look at? > > >> > 2) In the current spec, a topic is the same as a feed URL from the > >> > publisher. But could the hub aggregate and / or segregate content > >> > from incoming feed data and create its own topics? > > >> For my part, I would see that as a natural extension. How would such > >> topics be created and managed? > > >> alexis > > >> > (For example, in item 2, our hub would aggregate press releases and > >> > the segregate them based on industry / language / geography into > >> > multiple topics that users can subscribe to). > > >> > 3) Separately, we're also interested in being able to filter by tagged > >> > field values to create filtered feeds. > > >> > -Danny > > >> > On Jun 29, 7:44 pm, Brett Slatkin <[email protected]> wrote: > >> >> On Tue, Jun 29, 2010 at 4:35 PM, Alexis Richardson < > >> [email protected]> wrote: > >> >> > On Wed, Jun 30, 2010 at 12:31 AM, Brett Slatkin <[email protected]> > >> wrote: > >> >> >> On Tue, Jun 29, 2010 at 3:30 PM, Alexis Richardson > >> >> >> <[email protected]> wrote: > >> >> >>> On Tue, Jun 29, 2010 at 11:09 PM, Bob Wyman <[email protected]> wrote: > >> >> >>>> Now that we've got substantial experience with topic-based > >> PubSubHubbub, > > >> >> >>> With all due respect - I don't think we have enough yet. The spec > >> is > >> >> >>> still unstable. > > >> >> >> Specific wording aside, Alexis, it's a fine time to be talking about > >> >> >> these ideas, right? > > >> >> > I don't wish to diss Bob's push in this direction. But I'd like to > >> >> > see the topic model 'settle down' before looking at content. I don't > >> >> > believe we are there yet. > > >> >> Gotcha. That's fine. I think everyone wants things to move slowly. > >> >> This is the first I've seen of Bob's ideas in this direction and I've > >> >> got to let it marinate in my brain a bit. But I'd imagine some new > >> >> concerns about the existing spec may fall out of it, which is probably > >> >> a good thing, even if we all collectively decide that they should or > >> >> should not be in scope. > > >> >> -Brett
