[OSM-talk] Invisible coastline errors in Potlatch

Richard Fairhurst richard at systemeD.net
Tue Jan 6 14:57:40 GMT 2009

Marc Schütz wrote:

>> The solution is either to move splitting ways entirely to the server (so,
>> although the user hits 'split', the way isn't actually split until the
>> server returns a message), or to fix the server. I'd be interested to know
>> _why_ the server runs so slowly at certain times.
> ... or by fixing Potlatch to notify the user and suggest a retry?

Oh come on, if it was that simple I'd have done it already. :)

In fact, I have done. On 8th June, 2007, according to the changelog on  
the wiki. Since then, whenever Flash Player lets Potlatch know of such  
a failure, Potlatch does indeed alert the user and offer the  
opportunity to retry. Have you ever actually used Potlatch?

But if the ActionScript VM in Flash Player doesn't tell Potlatch that  
there's a problem with the connection, Potlatch can't know to tell the  
user anything.

It could, of course, go out of its way to poll the server and ask if a  
particular operation has been completed. Believe me, I've spent long  
hours trying to work out whether that's possible. In API 0.5 it's not  
really practical, unless you have some serverside storage to maintain  
a list of what operations have been requested and/or completed: about  
the only place you could do this would be user preferences, which  
would be horrid.

API 0.6 exposes version numbers, which makes it more feasible. But (a)  
polling the server isn't going to be that helpful anyway in times of  
high server load, (b) the previous query may just be running very,  
very slowly and will eventually execute anyway. IMX in a majority of  
cases, the second split write does execute, it's just that the server  
doesn't tell Potlatch about it.

Of course, if you have a better idea, feel free to expand on it in  
more words than "just fix it".


More information about the talk mailing list