[postgis-devel] Vote to drop support for 8.4 in 2.1

[Mark Cave-Ayland]

On 23/05/12 07:46, Paragon Corporation wrote:

> Now that we are all entrenched in 2.1, I'd like to get this out of the
> way so we don't have to waste time testing it anymore.
> I propose we drop 8.4 support in 2.1 because:
> 1) 2.1 is going to have to support 9.0,9.1,9.2 (and yikes maybe even 9.3)
> 2) There are many changes in how plpgsql works between 8.4 and the 9.0
> that makes it difficult to cross
> test - for example how aliases are handled in 9.0 -- so code that works
> in 8.4 may not work in 9.0
> 3) lack of ordered by aggs in 8.4 which I feel might be important later
> on for raster and topology
> 4) improvements in window functionality in 9.0 missing in 8.4 (e.g.
> great improvements in RANGE and intro of numbered ROWS)
> which again would be useful probably more so for topology (possibly
> tiger geocoder) than any other piece of PostGIS.
> 5) 9.1 and 9.2 will support KNN GIST, 9.2 will support SP GIST so we are
> already coding exceptions. I know we still have to for 9.0
> 6) PGXS - Mark's hack is needed for 8.4 but not for 9.0 etc. -- Mark can
> you confirm?

Yes that's correct - I wrote a comment about that in the override 
Makefile here: 
http://trac.osgeo.org/postgis/browser/trunk/postgis/Makefile.pgxs.

> 7) I think in 2.2 we'll probably be forced to drop 9.0 and dropping two
> versions in one release I think is too steep of a slope for our users.

As long as we are clear about what we do and do not support, i.e. we 
clearly fail at configure time if the installed library versions are not 
sufficient then that's fine with me.

+1.


ATB,

Mark.