[postgis-devel] RID Enforcement
Paul Ramsey
pramsey at refractions.net
Fri Feb 4 16:08:05 PST 2005
I think that would be a fabulous change.
It would allow Mapserver, among others, to do a bounding box query
without having to look up the srid, which means that a missing
geometry_columns entry would not cause things to blow up that were
otherwise correctly put together. Mind you we have 3 years of
installed base of mapserver which will continue to work the old way,
and would have to put a version check into a patched mapserver...
I wish we could just get geometry_columns to automagically update, but
sadly that remains out of reach.
P
On 4-Feb-05, at 1:27 AM, Markus Schaber wrote:
>
> Paul Ramsey schrieb:
>> It appears that findsrid is failing for reasons unclear. Oh, wait, do
>> you have valid entries in geometry_columns for your tables/geocolumns?
>>
>> I swear to god, one of the stupidest things we did in PostGIS was
>> enforce SRID coherence in queries. It is "for your own good" but the
>> extra maintenance overhead totally makes it not worthwhile. Better
>> just
>> to blindly assume the user knows what they are doing. This way it is
>> trivially easy to make a system that is broken for no obviously
>> visible
>> reason.
>
> Maybe the best compromise would be to make SRID -1 (no SRID) a special
> case. Allow operations between equal SRIDs, and between any SRID and
> -1.
>
> This should put away most of the annoyances, but still hinder the users
> to accidentally work with incompatible SRIDs (they still can reproject
> them if needed, or use setSRID() if they know what they do).
More information about the postgis-devel
mailing list