Yeah, the problem is that PubSubHubbub is predicated on the existence of a feed (either RSS or Atom). I think it would be workable for the API for public content, but otherwise I also see the authentication issue as a big problem. I'm pretty sure that some patterns are emerging for authenticated content using PubSubHubbub, but I don't think that aspect is very mature yet, so we're probably better off waiting another 6 months rather than trying to invent something ourselves.
Ethan On Fri, May 21, 2010 at 2:30 PM, Richard Hirsch <[email protected]>wrote: > On Fri, May 21, 2010 at 9:27 PM, Richard Hirsch <[email protected] > >wrote: > > > > > > > On Fri, May 21, 2010 at 9:23 PM, Ethan Jewett <[email protected]> > wrote: > > > >> We would be a subscriber, though we could conceivably set up actions or > >> something else that acts as a publisher as well. It's one of the options > >> for > >> really stream-ifying the API. > >> > > > > I think an action that publishes would be perfect. > > > > I was looking at the slides here (http://pubsubhubbub.appspot.com/) and > the > last slide has details for a publisher: > > > - Publish Atom feeds with their content > - Include Hub forwarding information in feeds (optional) > - POST updates to Subscribers and Hubs (optional) > > > I think an action by itself wouldn't be enough. I also don't know how > authentication might work with Atom. > > > >> This would only affect RSS/Atom actions. The way those actions currently > >> work is that they poll an RSS feed when they are triggered. We almost > >> always > >> set them up to run regularly using an "every 5 mins" test, for example. > >> PubSubHubbub gets rid of the need to poll at all, since the hub will > >> automatically notify the ESME instance every time the feed is updated. > >> > >> This would kind of change the semantics of the RSS/Atom actions from > being > >> an action to being more of a test, actually. Interesting. Not sure if > this > >> is what we want. Does anyone use RSS/Atom actions with a test other than > >> "every X mins"? > >> > > > > Doubt it > > > >> > >> Ethan > >> > >> On Fri, May 21, 2010 at 2:04 PM, Richard Hirsch <[email protected] > >> >wrote: > >> > >> > Sounds way cool - so if I understand PSHB correctly - we would then > be > >> a > >> > subscriber as well as a publisher? Why I don't completely understand > is > >> > the > >> > connection to actions? Would there be a PHSB action that sends > messages > >> to > >> > a > >> > hub? > >> > > >> > D. > >> > > >> > On Fri, May 21, 2010 at 7:36 PM, Ethan Jewett <[email protected]> > >> wrote: > >> > > >> > > Not very far, but enough to build on, and I plan to keep going once > >> I'm > >> > on > >> > > my sabbatical (starting in 1 week!!). There is a branch in the svn > >> > > repository that has all my progress in it. > >> > > > >> > > The basic approach was in to parts: > >> > > > >> > > 1. Add an indicator to the action that the feed is PSHB-enabled > >> > (determined > >> > > by analyzing the feed when the action is polled), and in this case > we > >> > would > >> > > never start the poller. Instead we would initiate a subscription. > >> > > > >> > > 2. The second half was to add a PSHB handler that could receive > >> callbacks > >> > > from hubs related to specific actions and post messages based on > these > >> > > callbacks. > >> > > > >> > > I think I got about 1/3 done with each of these tasks :-) > >> > > > >> > > An additional nice-to-have would be to implement use of the optional > >> > > signing > >> > > scheme for hubs that support it, though this is not necessary to > reach > >> > our > >> > > current level of security. > >> > > > >> > > Ethan > >> > > > >> > > On Fri, May 21, 2010 at 12:26 PM, Richard Hirsch < > >> [email protected] > >> > > >wrote: > >> > > > >> > > > @Ethan: Out of curiosity, how far did you get on your pubsubhubbub > >> > > > implementation? > >> > > > > >> > > > D. > >> > > > > >> > > > >> > > >> > > > > >
