Joerg Sonnenberger <joerg@britannica.bec.de> wrote:

> On Fri, Jul 15, 2005 at 09:20:14AM +0000, Valeriy E. Ushakov wrote:
>> 1) collation is sorting order, not toupper/lower mappings, so I don't
>>    understand how strcasecmp is related
> 
> (strcasecmp just like strcmp, for the sake of it)

Hmm, it seems to be poorly defined.  SUS doesn't even xref from
strcasecmp to strcoll and has no language parallel to the one used in
strcoll description, so it's not clear how it's realted to strcoll in
treating the ordering of "not-equal" case.

Anyway, strcoll demosntrates the point even more nicely, that's why I
was confused you mentioned strcasecmp.


> Never the less, since it is a (partial) order function, it can be
> represented as table or set of tables.

[offtopic alert]

I've been pondering the example with unicode-encoded tibetan, that, I
said, cannot be easily collated (strcoll) using only tables, b/c
lexicographical weights of characters are not 1-1 w.r.t the order of
characters in the stream.  But come to think of it, a table of
suitable size (few hundreds strings) can do the trick.

SY, Uwe
-- 
uwe@ptc.spbu.ru                         |       Zu Grunde kommen
http://www.ptc.spbu.ru/~uwe/            |       Ist zu Grunde gehen