[libvoikko] Using BCP 47 language tags in libvoikko

Harri Pitkänen hatapitk at iki.fi
Tue Apr 13 18:03:59 EEST 2010

On Tuesday 13 April 2010, Flammie Pirinen wrote:
> I might be reading the ABNF wrong-, but doesn't
>   privateuse    = "x" 1*("-" (1*8alphanum))
> mean that you could as well use fi-x-really-long-variant-name (or
> reallylo-ngvarian-tname assuming automatic mapping, of course)?

Yes, that seems to be the case. I misunderstood that in the specification.

> I haven't checked the algorithm, but I suppose that it can do
> something reasonable if you have e.g. only HFST variant and medical
> variant available. Of course in the end good user interface is always
> required.

There is a mechanism (mor-default symlink) already in use within libvoikko. I 
suppose we should still support that even if the standard algorithm specifies 
something else for these situations.

Or maybe not. Symlinks with multiple languages and variants can get quite 
complicated and the alternative for Windows (copying the entire directory just 
to specify the default) is ugly. There must be a way for user or administrator 
to select which variant is the default but the mechanism may need to be 
redesigned for dictionary format 3 (in libvoikko 3.1).


More information about the Libvoikko mailing list