[josm-dev] Conflicts because of timeouts from a proxy

Martin Vonwald imagic.osm at gmail.com
Sat Feb 23 10:19:19 UTC 2013


2013/2/22 Dirk Stöcker <openstreetmap at dstoecker.de>:
> On Fri, 22 Feb 2013, Martin Vonwald wrote:
>
>> What I now noticed is that right before the conflict appears JOSM
>> always (I think!) says something like "New try in x seconds". What I
>> suspect here is a timeout from the proxy I have to use at this place
>> here. JOSM then tries to resend the data but the server already
>> accepted it and therefore the resending creates a conflict.
>>
>> I have no possibility to test this assumption - at least I'm not aware
>> of one. Can anyone else test this? Please let me know if I can help in
>> any way.
>
>
> Sadly you have one of the cases we can't easily solve. The upload of data
> can take a relatively long time due to server data validation. Now your
> proxy closes the connection and JOSM does not know whether data is accepted
> or not (the OSM protocol does not have transaction security like e.g. SQL
> databases).
>
> There are two ways you can reduce such problems:
> a) Use the possibility to upload smaller changesets. This will reduce the
> validation time and hopefully prevent connection aborts.

Here's the problem: I can't make the changeset smaller. It is usually
just the split of one single way.


> b) When you had an abort, after some time (e.g. 2 minutes) do an update of
> the data (file menu). This should detect most issues and resolve them.
> Remaining issues should be duplicates of new entries. Here a validator run
> to find duplicates can help.

I found an easy solution that works for me. As the internet connection
is very fast and rockstable I can safely assume that the server got
all my data. Now I hope that there was no real conflict and simply
delete my data layer and redownload. Everything's there.
This of course will fail as soon as there is a real conflict.


> Until we get a better upload protocol or someone implements a clever "take
> my last changesets and test if current changes match" algorithm you need to
> live with that issue.

I'm aware of that now.

regards,
Martin



More information about the josm-dev mailing list