🚀 go-pugleaf

In article <8496@charon.cwi.nl> dik@cwi.nl (Dik T. Winter) writes:
>In article <1i2durINN2pj@rodan.UU.NET> avg@rodan.UU.NET (Vadim Antonov) writes:
  ....

> > >Moreover, one question: how would you encode the German A-umlaut such that
> > >it sorts properly (i.e. as if it is the letter combination AE)?
> > 
> > The sorting order should be strict -- if you have two identical words
> > with a-umlaut and ae in the middle is it the same word? If it is then
> > ae IS a variation of a-umlaut and should always be treated as a signle
> > letter.
> > 
>I do not think you understand.  From the AVON (Amtliches Verzeichnis der
>Ortnetzkennzahlen) edition 1985, which gives area codes for the places in
>Germany.  The next is a selection of places mentioned ("o is o-umlaut):
>	Modautal
>	M"ockm"uhl
>	...
>	M"ornsheim
>	Moers
>	M"ossingen
>	...
>	M"otzingen
>	Mogendorf
>now come up with a coding that allows this (standard German) sorting.
 [and further mentions that "o is equiv. to oe, but oe is not equiv. to
  "o.]

     I'm afraid that Dik Winter, as well-informed as he is, has
     oversimplified things a little. :-)  There is not one sorting
     scheme used in Germany, but (at least?) two.

     The scheme described by Dik *is* the scheme that is used in
     telephone books, (and generally for proper names?) for very good
     reasons.

     It is *not* the sorting that is used in any German dictionary I
     have seen.  In dictionaries, a-umlaut is equivalent to a (not ae)
     for sorting purposes, o-umlaut to o, u-umlaut to u, and "sharp s"
     to ss.

     This further complicates Vadim Antonov's scheme for having character
     codes encode sorting order: we would have to somehow specify that
     under certain circumstances  "a  sorts one way, and under others
     it sorts another.

     I even seem to recall a third scheme, but I don't remember what
     it was, and it could have been something only used by users
     of English-"speaking" computers.


     (However, I am inclined to think that totally different alphabets,
     e.g., the Cyrillics and the Latins, should be represented separately.
     I am not familiar with 10646, but some posts here have suggested
     that 10646 uses the same code for all letters that look like, e.g.,
     T -- I can understand someone not being happy with that, if it is
     true.  However, I don't have much experience in producing a unified
     character set.)


-- 
Alan McKenney        E-mail:  mckenney@cims.nyu.edu         (INTERNET)
Courant Institute,NYU,USA     ...!cmcl2!cims.nyu.edu!mckenney   (UUCP)