[OSM-talk] Re: [OSM-dev] MassGIS dataset

[SteveC]

* @ 04/06/06 05:26:26 PM crschmidt at crschmidt.net wrote:
> On Sun, Jun 04, 2006 at 05:17:34PM +0100, SteveC wrote:
> > * @ 04/06/06 01:34:58 PM crschmidt at crschmidt.net wrote:
> > > What proof can I offer? What more can I do? Give me a case where you
> > > think there will be problems, and I'll do my best to see that there
> > > aren't. Without some idea of what you think will break, I can't convince
> > > you otherwise, and neither can anyone else. We've had three people with
> > > experience in databases looking at the way OSM does things, offering
> > > advice, or offering to help improve things... but no one other than you
> > > understands your concerns.
> > 
> > What I havn't seen are tested patches.
> 
> You're right. I'm sorry I don't know how to test such a thing in this
> specific case.

With the node API calls specifically the only non-obvious thing is when
one API call is updating a node (moving, deleting) or creating one then
a getnodes call or any getnodes derivative like the 'map' api call
could get consistent data. Specifically without some kind of locking
then you could get

query 1 part 1: update all the old node X's to be deleted
query 2: get all the nodes in some area (which include X)
query 1 part 2: insert the new node X row information

So query 2 would fail to get node X.

The mysql downtime for creating the new field etc is a necessary evil...

Then the same concept can be expanded to the segment and ways tables.

have fun,

SteveC steve at asklater.com http://www.asklater.com/steve/