On Fri, Jul 02, 2004 at 06:31:47AM +0100, Peter Karlsson wrote: > > > This is quite counter-productive, since some version changes in the > > english wml files are sometimes just cosmetic and the translation was > > quite alright when it was removed. > > And sometimes the changes are quite extensive. How is the script > supposed to be able to tell the difference, without empowering it with > some kind of artificial intelligence? There's no need to empower it with AI in order for it to be able to detect, for example, that only a single line was changed between the version X (translated at some point) and version Y (current). 'diffstat' might not be a perfect metric (specially if the write rewrapped lines without making any changes) but it might be sufficient as a first attempt to avoid removing translations that have not kept up with very simple changes in their original versions. Granted, you can change a single line in a text in a way that makes the text itself completely different from a previous version, but for most common cases (typo fixes, additions, removal of text) it might be good enough or, even, sufficient to rule out that the translation is completely out of date both in a "version comparison" metric ( X << Y ) and in a "content comparsion" (insertions+deletions in 'diff -u X Y | diffstat' are greater than an arbitrary number) Just a few thoughts. Javier
Attachment:
signature.asc
Description: Digital signature