On Thu, Jan 14, 2010 at 6:15 AM, Craig Ringer
<craig@postnewspapers.com.au> wrote:
>>> Out of interest: Why not?
>>
>> There's plenty of discussion in the archives about it, but basically
>> ICU would represent a pretty enormous dependency and would lock us in
>> to having no other backend encoding but UTF8.
>
> Thanks. You're right - I should've just STFA ;-) so I appreciate the
> brief explanation.
There's also the question of whether being inconsistent with the rest
of the system is really the right thing at all. If a programmer does a
sort in the database and then writes application code using the same
collation on the same system which depends on the data being sorted
it's nice that that works. Or if an admin presorts the data using sort
before doing a data load he might reasonable expect the table to be
effectively clustered.
Switching to ICU means trading our current inconsistency from platform
to platform for a different inconsistency which would be better in
some cases and worse in others.
--
greg