On Wed, 2002-11-06 at 15:40, Joe Moore wrote: > Derek Gladding said: > > Hypothetical question... > Hypothetical answer below... > > > If one took a spell-checker, such as Aspell, then: > > > > - piped the whole of Usenet through it for a couple of weeks > > - automatically removed all sequences of characters that failed > > - removed all duplicate words from the result > > > > Would the resulting list of words be a "new" creation, unencumbered by > > any license attached to the spell-checker ? ;-) > > No, this would be a derivative work of the Usenet postings, which are > copyright their authors. > > You'd have to get permission from all Usenet posters.
It would not be a derivative work of Usenet, because the only elements copied would be uncopyrightable individual words. Usenet is, of course, a poor place to get content, because your wordlist would be filled with (a) foreign-language messages and spam, (b) made-up words like froup and kibo which (probably) shouldn't go into the dictionay. I've done some research into this issue with one of the attorneys who works with us. We feel that the following is the best way to proceed: Take the word list from gcide and the existing aspell-en list. diff gcide aspell-en|grep '^>'|cut -d ' ' -f 2 > missing-words cat missing-words gcide|sort So, we're starting from gcide, which is Free. Then we're seeing what we're missing from the old aspell-en list, and adding it in to the gcide list. The process is, in other words, as important as the product. It's pretty doubtful that word lists have copyright -- that they are closer to phone books than novels. We feel that there's unlikely to be trouble of any sort related to this use of the old aspell-en list. Since we're not directly distributing the aspell-en list, we're even more likely to be OK. I'm CCing RMS here because I know he's interested in hearing the tentative decision we've come to. -- -Dave Turner GPL Compliance Engineer Support my work: http://svcs.affero.net/rm.php?r=novalis&p=FSF