En réponse à Philippe Caquant :

>I'm on Windows XP Home.
>I noticed that on the Web, you can find "ranges" of
>Unicode characters to download, as single fonts.

Unicode is huge. Making a single font containing all the characters is just 
way too much work. There are fonts trying to do that, but they are not free 
(Arial MS Unicode, which is about 26MB large, used to be free, but not 
anymore...) and they never really fill all the ranges.

>  I
>first thought I just had to load one single Unicode
>font and than all glyphs would be at hand when using
>the function "Insert / Special Characters" in Word, be
>it Tibetan, Amerindian or Braille. It seems like it
>doesn't work this way.

You are extremely naive. Practically, we're already used to using various 
fonts for various parts of the same document (for instance Arial for titles 
and Times for text). What's wrong in using various fonts for various 
scripts? the only thing Unicode ensures is that the fonts will be 
compatible, i.e. that no font can have a Cyrillic letter where another font 
has a Japanese katakana.

>  So I don't know exactly what I
>should do.

Download fonts containing the ranges you need.

>  At the moment, the Insert function just
>proposes me some Latin, Cyrillic, Arabic, Hebraic,
>Symbol codes (does this mean that other codes weren't
>at hand in release 2.1, or just that the font
>currently used is a subset of Unicode 2.1 ?) and
>that's about all.

What font is it? The Insert works on a per font basis. Look at other fonts 
and see what ranges they propose. If you don't have the ranges you need, 
look on Internet and download the fonts you need.

>  So I suppose I should load all
>range-fonts one by one (how many are they ?),

That's a nonsense question. There's no body that created a set of fonts 
filling all the ranges in Unicode. Fonts are created by people who need 
them, and they put in the ranges they need. There's no such thing as a 
"range-font". Just look for fonts containing the ranges you need and 
download them. They will always overlap in some ranges.

>  and
>before doing an Insert / Special character, changing
>to the font I would like to use. In that case, of
>course the macros would be much more complicated to
>write, because you had to add such tests as (for
>- if the code is in the range[x,y], than first change
>to font F, supposing font Z is at hand
>- if it's in the range[x',y'], then tell the user, one
>way or another, that he forgot to install font F'
>- etc.

That's what Unicode capable programs already do, in reading. I look at my 
browser, Opera, for instance. If a character it has to show doesn't appear 
in the main font of the webpage, it will use the character from another 
font to show it. If I have no font containing it, it will show a "no 
character found" character (depending on the encoding, it can be an empty 
rectangle or a diamond with a question mark in). Of course, if you're using 
Internet Explorer, you don't have this chance. IE isn't able to pick 
characters from other fonts than the one the webpage is written in. That 
makes it very Unicode-unfriendly.

Is it so difficult to understand that the macro you're talking about is 
unneeded? Just use modern Unicode-aware tools, and it will already be built 
in. But since not everyone on this list can use modern tools (for various 
reasons), don't use Unicode on the mailing list. People who could use the 
macro you're talking about won't need it, because they already have 
Unicode-aware tools, and those who cannot use Unicode-aware tools wouldn't 
be able to use your macro anyway. Why reiventing the wheel?

>Then probably the different fonts wouldn't belong to
>the same release, etc.

As stated already, Unicode has been stable since 1996. Since then, it has 
only *added* characters. It hasn't *changed* any character that was already 
there. Do you have any font that is more than 8 years old? I doubt it. So 
it means that the only thing will be that your older fonts will contain 
less characters than the new ones. But those characters will be at the 
place where they are expected to be. In other words, all the fonts will be 

Christophe Grandsire.

You need a straight mind to invent a twisted conlang.