On Sun, Jan 31, 2010 at 4:02 PM, Blaine Cook <[email protected]> wrote:
> On 31 January 2010 22:56, Michael Barinek <[email protected]> wrote:
>> I'm not sure if location based "topics/feeds" have been discussed in the
>> past, but here's something else that I've been playing around with.
>>
>> I've been including 'hub.lat' and 'hub.lon' parameters in subscription
>> requests and then filtering inbound feeds by location (when available).
>>
>> for example...
>>
>> hub.callback=http://your.hostname.com&hub.lat=39.51&hub.lon=76.24
>>
>> I'd be interested in getting feedback - interesting, not interesting,
>> doesn't fit...type stuff?
>>
>> Here's a small application spike (still rough around the edges) that I've
>> been playing around with http://www.localhash.com/
>
> This sort of thing is definitely what I'd hoped would happen with
> PubSubHubbub, and have been advocating in the context of XMPP PubSub
> for some time! Yay!
>
> On the subscriber side, it's not really necessary to encode the
> lat/lon into the URL, since you really should be keeping a record of
> the subscription locally (if not, you're vulnerable to a number of
> attacks that are much easier to manage with a subscription record).
> Instead, store the filters internally and do the filtering as you
> would normally, possibly including an opaque identifier as a pointer
> to the filter.
>
> If possible, a more robust and efficient approach would be for the
> service provider to offer feeds (i.e., topic URLs) that accept the
> lat/lon parameters as part of the subscription. So, instead of
> subscribing to:
>
> "http://geoprovider.com/feed.atom";
>
> you'd subscribe to:
>
> "http://geoprovider.com/feed.atom?lat=39.51&lon=76.24";
>
> and the provider (or, in many cases, the hub by virtue of fetching the
> provider's feed) would take the responsibility of only sending
> relevant atom entries.

And the huge benefit of this approach is that the PubSubHubbub
protocol does not need to change at all to support this-- you only
need a hub/publisher that's intelligent enough to interpret the
subscription URLs properly.

Reply via email to