Sorting compound names

[jda]

>I've dealt with this before, and the further I delved into it the 
>more I came to realize that the problem cannot be solved 
>computationally at all.  Well, I should qualify that first by 
>agreeing that it might be possible in some culturally and 
>linguistically "pure" backwater somewhere that rules and customs for 
>family names were consistent enough to be codified.  But as soon as 
>you have multiple countries, cultures, languages, and even (yes!) 
>the preferences of the individuals involved -- it becomes absolutely 
>impossible to codify.
>
>The only real solution is to have a separate field for full-name 
>sort-order, and then populate it manually -- or programatically -- 
>but either way being prepared to make lots of corrections if people 
>complain.
>
>I've compiled a short list here of names, sorted "properly" 
>according to examples in the _Chicago Manual of Style_ (entries 
>18.102 - 18.107).  I'm not in any way claiming that this is a 
>definitive authority -- I'm merely suggesting that this is a *large* 
>can of worms that you probably want to avoid opening if at all 
>possible.  Again, this list is "correctly" sorted:
>

Hi Erik,

Thank you very much for your answer.

I already have users complaining, however. One wrote

"This is one of these anglocentric idiosyncrasies which horrify 
Germans, Dutch, Swedes and so on."

    :->

Well, I didn't point out English is not the only language that uses 
strict alphabetical sorts (and that it should be "that horrify", not 
"which horrify", but...).

I see that EndNote allows users to specify words that are *not* 
included in name sorts (things like "van", "von", etc.). They also 
let you specify words that should be ignored at the beginning of a 
title for sorts (words like "The", "A", "An", etc.).

This seems to be the sort of thing that would be acceptable. I can of 
course get the names in each record, strip out the specified "words", 
and then do a sort (in RB), but this will obviously take a 
substantial amount of effort and be much slower than a method in 
Valentina. So, I'm still wondering if it is possible to construct 
such a method (I'm posting to the betas list, BTW, because I gather 
that methods will be more flexible in 2.x).

Thanks again,

Jon