[postgis-devel] Proposal to drop support for PostgreSQL 8.3 in PostGIS 2.0

[Mark Cave-Ayland]

Paragon Corporation wrote:

> While we are on the topic of breaking changes, I propose we drop support for
> PostgreSQL 8.3 in PostGIS 2.0 and make our further changes with that in
> mind.
> 
> Reasons:
> 
> 1) As a rule of thumb, I don't think we should support more than 3 versions
> of PostgreSQL in any major/minor release.  Its just too many permutations of
> testing involved.  Its pretty clear to me we will have to support
> 
> At least PostgreSQL 8.4, 9.0, and upcoming 9.1
> 
> 2) If we drop PostgreSQL 8.3, we can start using the default parameters
> features for functions introduced in 8.4.  I think this will get rid of 30%
> of our .sql function code base e.g. ST_AsKML, ST_ASGML etc. can be reduced
> down significantly.  Many of the raster functions as well.  We have a whole
> slew of such other things.
> 
> I know this may require document changes, but would be willing to go thru
> that pain to get reduce our 700 someodd function permutations.
> 
> 3) We can probably get rid of that change Mark put in for installing that
> says -- Get rid of this when PostgreSQL 8.4 is the minimum.  Not to mention
> several other places I can only guess we have in place just to support
> PostgreSQL 8.3
> 
> 4) If we don't drop support for PostgreSQL 8.3, I fear in 2.1 we might have
> to drop support for 2 versions of PostgreSQL.  I feel from a user standpoint
> gradually dropping off versions is less painful than dropping 2 versions.
> 
> Just my 2 cents,
> Regina

 From my point of view it's dependent upon the overhead placed on 
developers, i.e. by supporting all these extra versions are we making 
more work for ourselves? In terms of function signatures, 99% of the 
time is spent writing the function while 1% is pasting the prototypes so 
for me it's a reasonably insignificant part of the process.

In terms of supporting versions, I'm generally quite happy to keep 
compatibility up to the point where we need a specific minimum version 
unless it causes a giant fork in the code. Then again, I appreciate that 
building Win32 installers can take a lot of time.

My gut feeling is that we should leave the supported PostgreSQL versions 
where they are for the moment, but I'd have no objection if you only 
built Win32 binaries for the latest 2 releases. I know I did this when I 
built them, but then again since StackBuilder is so easy to use then 
generally Win32 users have the easiest upgrade path and so it is 
reasonably painless.


ATB,

Mark.

-- 
Mark Cave-Ayland - Senior Technical Architect
PostgreSQL - PostGIS
Sirius Corporation plc - control through freedom
http://www.siriusit.co.uk
t: +44 870 608 0063

Sirius Labs: http://www.siriusit.co.uk/labs