Joshua, et. al.
Sorry for weighing in on this discussion so late; I've gotten behind on the
Hackers digests, since there are some 90 messages a day.
I don't see anything wrong with the idea of maintaining a 7.3 tree for bug
fixes and testing if /contrib modules can be backported. Heck, I'm pretty
sure that tsearch, pgavd, and erserver already have 7.3 compatible downloads
on Gborg.
What I would resist is the idea that any contributors be distracted from work
on new+improved features for 7.4 & 7.5 for this. That is, whoever maintains
the 7.3 tree should be someone new, or someone who's not up to hacking new
features, yet.
Actually, this might be a great way for a new hacker to get up to speed on the
PostgreSQL codebase.
Comprehensive backward compatibility is the hobgoblin of commercial software
development. I support one legal services database (Bruce knows who I mean)
which after 10 years of development still sport the same ill-concieved and
poorly normalized schema, which leads them to constant serious performance
and scalability issues. Why? Because 80% of their customers still use 2-6
year old versions of their software, and thus their development team spends
75% or more of its time supporting old versions instead of creating new ones,
and they're not allowed to re-structure the schema becuase that makes
upgrades more costly.
We don't want to go this way.
So: yes to keeping a 7.3 tree; No to having Bruce, Tom, Joe, Neil, Sean,
Peter, Alvero, Jan, or any other current major contributor bothered with it.
(Hmmm ... don't we have even one woman on the contributors list? Geeks are
us)
--
Josh Berkus
Aglio Database Solutions
San Francisco