If you decide to build your own "hub" software, the publisher -> hub
relationship is really up to you.
At superfeedr for example, we implement the regular spec for this part, but
also allow other protocols (XMPP, XML-RPC) or event fat pings.
What matters the most is that your "compliant" on the hub -> subscriber
part, as long as you run your own hub.


On Thu, Apr 28, 2011 at 7:44 PM, Justin Karneges
<[email protected]>wrote:

> Indeed.  My thought is that the feed could know this by means other
> than the mechanism described in section 7.1.  For example, if AtomPub
> was used to create an entry, then this could be enough for the "hub"
> to push an update out to subscribers.
>
> On Apr 28, 12:54 am, Julien Genestoux <[email protected]>
> wrote:
> > Hey Justin,
> >
> > I don't see any reason why that wouldn't work.
> > However, your "hub" will still need to know when the feed has been
> updated
> > to push the updates to the subscribers.
> >
> > Cheers,
> >
> > On Thu, Apr 28, 2011 at 4:45 AM, Justin Karneges
> > <[email protected]>wrote:
> >
> > > Hi folks,
> >
> > > We would like to offer push callbacks to our feeds without the notion
> > > of a separately addressable hub.  So, for example, if we have an Atom
> > > feed here:
> >
> > >  http://example.com/feed/
> >
> > > Then, one could subscribe to updates using the following:
> >
> > >  POSThttp://example.com/feed/
> > >  params:
> > >    hub.callback={callback_url}
> > >    hub.mode=subscribe
> > >    hub.topic=http://example.com/feed/
> > >    hub.verify=async
> >
> > > Of course, the topic URL must always be the feed itself.  The "hub"
> > > would never crawl out to some external feed, nor would publishers
> > > necessarily need to notify the "hub" of changes, since the feed and
> > > hub are one.
> >
> > > Is there an example of this approach in the wild?  Is this approach
> > > discouraged in any way?
> >
> > > Thanks,
> > > Justin
>

Reply via email to