Thanks for the clarification, Martin.
Our goal has always been to be able to support as many languages as possible
on all of our supported platforms. There are currently two approaches that
can be taken to accomplish this:
- make ispell Just Work
- make pspell Just Work
They each have their advantages and disadvantages.
1. make ispell Just Work
-------------------------
To be clear, nobody chooses ispell for technical merits any more. We
originally made a pragmatic choice to integrate a forked version of Ispell
because it already had dictionaries for 16 of the 25 locales that we
currently support, plus an additional 9 locales that we don't.
The ispell codebase is, admittedly, creaky old code that gives Unix a bad
name. However, at the time it was our best alternative to allow us to get
spell checking that Just Works, so we held our noses and various heroes have
volunteered to dive into that codebase to:
- port it to Win32 and other platforms (Eric et al)
- deal with endianness issues (Darren)
- deal with some non-Latin-1 issues (Henrik)
- etc.
We currently have at least one problem that's blocking ispell usage in
certain situations. These haven't been well-identified, and thus nobody's
stepped up to fix them. I continue to suspect that once the current problem
is isolated, the necessary fix should be easily localized, but I haven't dug
in enough to find out.
2. make pspell Just Work
-------------------------
As Martin points out, there are a lot of technical advantages to the pspell
approach, plus a very motivated developer (hi, Kevin!) who's actively moving
that codebase forward.
To date, nobody's stepped forward to investigate the work needed to get this
working for our various non-Linux platforms, and I'm not clear what the
coverage of existing dictionaries is.
bottom line
-----------
I've been on the record for quite a long time that I'd love to have a better
spelling solution than we currently have. For a list of criteria, see:
http://www.abisource.com/mailinglists/abiword-dev/00/February/0457.html
Any solution (based on ispell, pspell, or some other engine) which does a
better job of meeting those criteria is worthwhile to have. We each may
have our own personal opinions about which alternative is preferable, but
the bottom line is clear:
whoever writes the best code wins
Anyone who feels motivated to pursue *either* approach should feel free to
go ahead and volunteer to do so. We want this problem resolved, period.
Paul,
spoiled by his en-US locale
