[OSM-dev] Reading OSM History dumps

Scott Crosby scrosby at cs.rice.edu
Mon Aug 23 15:29:33 BST 2010


On Mon, Aug 23, 2010 at 8:39 AM, Frederik Ramm <frederik at remote.org> wrote:
> Hi,
>
> Scott Crosby wrote:
>>
>> My idea is that we assume that a given snapshot of the database at a
>> given changeset number is unique.
>
> I don't know if that breaks your whole edifice but the above assumption is
> false; changesets are not generally applied in a transaction so at any one
> time, hundreds of changesets may be open and may, during the course of their
> lifetime, modify the same objects.

Non-atomic changesets are both a fatal problem in designing and
constructing a history and at the same time, irrelevant.

Creating a true history is impossible. The information about the
concurrent updates across different non-atomic changesets is
presumably lost.

This is not actually a problem. To present a history, all that is
needed is to define *an* ordering that is usefully consistent with the
OSM database history. Pretending changsets are atomic lets me
define/construct one such ordering, using changeset numbers or ending
timestamps. Peter's design seems to make a similar assumption about
atomicity of changesets.  Other orderings, such ignoring changeset
numbers entirely and using timestamps of entity changes would have
their own tradeoffs.

Scott



More information about the dev mailing list