[OSM-dev] Unexistant wayed segments in planet?

Tom Hughes tom at compton.nu
Sun Sep 16 23:54:49 BST 2007


In message <46EDAEAF.5020904 at bretth.com>
          Brett Henderson <brett at bretth.com> wrote:

> Tom Hughes wrote:
>
> > Not as far as I know it won't...
> >
> > Well it could if you used the task that reads the history tables but
> > that is (a) slower and (b) does not produce consistent results. Hence
> > the new task that was added to use the current data tables. That task
> > cannot produce a consistent snapshot however.
>
> By non-consistent results do you mean that the old TIGER data is
> included?  If so I assumed this was a temporary problem that could be
> addressed if/when current database issues are resolved.  I'm sure there
> will be other minor discrepancies due to the lack of transactional
> database support across multiple db updates but these should be
> relatively few and easily corrected via the API.  The way history
> problem will go away nicely when API 0.5 is deployed, so that only
> leaves nodes.

Well it's non-consistent in as much as there is no guarantee that
every change is successfully recorded (and has always been recorded
in the past) in the history tables.

Of course 99.999% of changes will have been recorded fine but without
proper transactions there are bound to be some changes that have not
preserved integrity properly.

Whether we ever decide to fix up the old Tiger data is another question
all on it's own of course.

> As far as speed is concerned, it should be much faster.  Once a
> consistent snapshot is achieved from that point on you can use the
> changeset derivation task to only pull changes within a specific time
> interval.  This should drastically reduce the load on the primary
> database and perhaps even allow more frequent data extracts.

That is a very good point.

Tom

-- 
Tom Hughes (tom at compton.nu)
http://www.compton.nu/




More information about the dev mailing list