PostgreSQL 8.2 has reached it's official end of life, meaning that now 8.4 should, I think, be the minimum version we should be targetting for any and all new development (including add-ons). There are a number of cases where performance is currently limited by us trying to support PostgreSQL 8.3, and the biggest example here is the rendering of the menu.
Regarding the menu, we decided out of band that we'd create a FAQ item telling people to load the Menu.sql file from trunk - which we promise to keep fully backward compatible - if they're on 8.4 and suffering the performance issue. So that part is "handled".
Myself I thought we required 8.4 for 1.3, but as it turns out both INSTALL and COMPATIBILITY document that we don't support 8.2 and before (each in their own way). So, I guess we're indeed stuck with 8.3 for now. I wonder if we can't make Menu.sql load conditionally depending on the target system: 8.3/8.4+ especially because the upgrade process will overwrite any trunk version loaded during patch release upgrades. We already have version detection in
setup.pl; all we need to do is load Menu-84plus.sql when we detect 8.4 or higher :-)
9.2 is on its way soon, and has a few features we will probably find extremely useful as we get there. In particular the native JSON support might be useful as a replacement for native Pg representation of nested types. I don't see that we can even think about requiring 9.2 for 1.4, however, since it will be in beta most of the time we are developing. However, I would like to consider looking at this for 1.5 despite the fact that it is a big jump.
Yea. It would be good to know which distributions (and their versions) offer which postgresql verisons: that'd allow easier decision making. However, for now, I think we're focussed 1.4. Based on the fact that 1.4 isn't there yet (we still have quite a bit of work to be done) and the upcoming 9.2 release, 8.3 will probably reach end of life soon after we release 1.4. If that's true, I think we should use that as a reason to not support it. Having at least another year of supported life is a good criterion IMO.
The reason here is that parsing/formatting complex data types for PostgreSQL is a pain, and currently DBD::Pg is not able to pick this up. We could probably standardize on XML at present, but JSON might be better long-run.
What do people think? Obviously any discussion right now needs to be pretty tentative.
I think supporting 8.3 - while noting the performance issue prominently - is a good thing for now, especially if we can ship Menu-84plus.sql (ie the trunk version of Menu.sql). Dropping it in due course also sounds like a good option - in a next minor release that is.
Bye,
Erik.