unifying spelling and dictionaries

[skierpage]

bug #5009 requests Hunspell "as it is the only free spellchecker that support features needed for a good Arabic spell checker".
Ideally OLPC would have one spelling engine, one extensible set of dictionaries, and a common dictionary for user-added words.

XULRunner thus Browse uses hunspell, I believe using dictionaries in /usr/lib/xulrunner-nnn/dictionaries. I think Abiword thus Write uses the enchant library which in OLPC is currently configured for aspell and myspell, not hunspell. I don't know what other XO apps use.

I think getting enchant and XULRunner to use the same hunspell dictionaries is "just configuration".

http://www.redhat.com/archives/rhl-devel-list/2006-September/msg00225.html is a useful thread on unifying.

[nemeth]

Source code of Enchant 1.3 contains Hunspell 1.4 under the name of "myspell". Really, it would be fine to use system wide spell checking with newer Hunspell version on XO, instead of using different Hunspell versions with parallel loading of the same dictionary (for example, single loading of en_US dictionary needs ~3-4 MB RAM). Last Hunspell supports gzip-like dictionary compression, saving MBs of disc space.

Fix the dictionary proliferation problem of Fedora: http://fedoraproject.org/wiki/Releases/FeatureDictionary

[khaled]

Since we are now based in fedora 9, I think this should be easily done since fedora did almost all needed job unifying spelling dictionaries to hunspell.