Re: ICU integration

On 31 August 2016 at 10:46, Peter Eisentraut
<peter(dot)eisentraut(at)2ndquadrant(dot)com> wrote:
> Here is a patch I've been working on to allow the use of ICU for sorting
> and other locale things.

Great to see you working on this. We've got some icky head-in-the-sand
issues in this area when it comes to OS upgrades, cross-OS-release
replication, etc, and having more control over our locale handling
would be nice. It should also get rid of the Windows-vs-*nix locale
naming wart.

Speaking of which, have you had a chance to try it on Windows yet? If
not, I'm happy to offer some pointers and a working test env.

How stable are the UCU locales? Most importantly, does ICU offer any
way to "pin" a locale version, so we can say "we want de_DE as it was
in ICU 4.6" and get consistent behaviour when the user sets up a
replica on some other system with ICU 4.8? Even if the German
government has changed its mind (again) about some details of the
language and 4.8 knows about the changes but 4.4 doesn't?

Otherwise we'll just have a new version of the same problem when it
comes to replication and upgrades. User upgrades ICU or replicates to
host with newer ICU and the data in indexes no longer correctly
reflects the runtime.

Even if ICU doesn't help solve this problem it's still valuable. I
just think it's something to think about.

We could always bundle a specific ICU version, but I know how well
that'd go down with distributors. They'd just ignore us and unbundle
it then complain about it. Not wholly without reason either; they
don't want to push security updates and bug fixes for bundled
libraries. Also, ICU isn't exactly a pocket-sized library we can stash
in some 3rdpty_libs/ dir .

--
Craig Ringer http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services