[OSM-dev] Removing Minutely and Hourly Changesets

[Kai Krueger]

On 01/-10/-28163 08:59 PM, Bernhard zwischenbrugger wrote:
> hi
>>> If all the changesets would go to an xmpp stream, it would be a
>>> really easy to such a liveticker for osm.
>>
>> There was a site in the past where this was visible. I have created an
>> OpenLayers implementation to visualize it. So even with what is publish
>> now, not a line of XMPP is required to do so.
> I did the osmlive based on minutely Changesets.
> The best result was about 5 minutes delay.
> On a webbrowser the xmpp lib strophe offers a nice BOSH connection and
> no polling is needed.
> http://code.stanziq.com/strophe/
>
> A 5 second delay could be done without problem.
> 5 seconds is faster than 5 minutes.
>
> Maybe you think 5 minutes are no problem.
> But think about an interface where you see live what other
> people are doing. The next step could be to implement a
> fast human to human communication.
> If you see somebody editing something in a strange way, you could give
> help on how to do it better.

With LiveEditMapViewerJ[1,2], which was heavily modeled after osmlive, 
can sort of do that. You can zoom into edits (currently only node 
changes) that look odd, shift click on one of the nodes to get to the 
OSM-Browse-page for that node, with the details of the changeset and 
editor and contact the user through the osm messaging system.

So far, I have thought the replicate diffs work really well for that and 
are probably even simpler than the old minutely diffs. You simply get 
the state.txt, which contains the most recent sequence number, and then 
you just count up by one every time you successfully retrieve the 
corresponding diff file.

With the replicate diffs, the delay is now down to normally at max a 
minute and on average probably more like 30 seconds delay. It wouldn't 
be hard to move the replicate diffs to be generated every 30 seconds, so 
on average delays of only 15 seconds.

Can you give examples for when that would not be enough? Given that the 
standard map call from the api often takes more than 30 seconds to 
complete, you will partly not get any more up to date data from the live 
API either.

Kai

>
> Bernhard
>
>
>

[1] http://gpsmid.sourceforge.net/misc/LiveEditMapViewerJ.jar
[2] 
http://trac.openstreetmap.org/browser/applications/utils/LiveEditMapViewer