to Mr. Peter Jacobi,
i contact to Mr.Theppitak Karoonboonyanan the man who
wrote http://www.nectec.or.th/it-standards/thaistd.pdf,
he has a full-mailbox problem.
his advice is contact him on threaded-web board at
this thread follow the post

http://linux.thai.net/wwwboard-devel/msg04163.php

**
now i use single byte 8 bit thai character .
i use postgresql7.3.4 on redhat 8 with locale th_TH (not th_TH.utf8).
i am migrating to firebird due to firebird already has win32 port,
i use linux and xp mixed in both server,workstation and stand alone.

thanks you.
somphong

** he is the man who work alot on thai opensource

--- In firebird-support@yahoogroups.com, "peter_jacobi.rm"
<peter_jacobi@g...> wrote:

> Hi somphong,
>
> --- In firebird-support@yahoogroups.com, "sp2gui" <sp2gui@y...> wrote:
> > i appreciate to be a tester.
> > and i will ask for other people to test it too.
> > thank you for your kindly and quickly response.
>
> May I ask you to have an actual look into
> http://www.nectec.or.th/it-standards/thaistd.pdf,
> especially the chapter about sorting?
>
> The doc subscribes a carelully laid out algorithm for
> dictionary sorting, using a full Unicode four level
> sort (if this is all greek to you, perhaps you can
> contact a local expert).
>
> From experiences with other locales, I can report the
> 'normal' users often find the sorting designed by
> 'experts' somewhat un-intuitive and would like another
> sort order. If you have any specific comment, please
> be bring it forward.
>
> Another implementatiuon decision to be made, is the
> definition of 'character'. What makes up one character
> will obviously influence the meaning of field lengths
> (as in 'CHAR(4)'), what can be matched against a single
> character wildcard in LIKE, the operation of the SUBSTR
> operator and some other minor points.
>
> Perhaps this decision is so obvious to native speakers of
> Thai, that it is not addressed thaistd.pdf? The easiest,
> pragmatic solution is of course one character = one byte,
> which will also gives the best performance. Another natural
> choice seems to be the 'cell' as mentioned in '3 Input Methods',
> i.e. base charcter + upper or lower vowel + diacritic or tone marker,
> the same unit which will deleted in whole when pressing the
> "Delete" key.
>
> Regards,
> Peter Jacobi