From: Matthew B. <mat...@co...> - 2005-05-24 14:42:09
|
I'm not saying Alistair Young wrote: >> What problems do you envisage? > > outside of classpath I have to worry about windows - do I have time? no! :-) >> Does this make it harder for people to be working on translations at >> the same time as someone has to merge the files back together > > not sure yet - tmx files are designed to be used with translation Properties files are also designed to be loaded into translation software and they are probably supported by more translation software than TMX. > applications rather than hand editors. Anyway, that's how cvs works. I > would think that users would donate translations to the community and we > would be responsible for merging them into the tree, just as we are for > code. Seems to be making more work for us. >> What do we gain? > > each template has only one resource file rather than, say 10 How is that better? >> What is wrong with the properties file approach? > > the biggest problem I've found is the fact that properties files are > meant for strings, not content. You can't have linebreaks in a > properties file value. As most web devs put these in to make it easier > to design, they'd have to remove them to localise the template. XML gets > round this problem. HTML translates all tabs, spaces and newlines into a space (unless inside a <pre>, I forget the CSS equivelent). >> Doesn't this mean someone has to have an understanding of XML > > not necessarily - we must stop thinking as developers and start thinking > as translators, who use gui apps and reuse localisations. I find the > lack of translator input into bodington very frustrating. Did most of the moodle translation get some by professional translators? I'm guessing it was some french bloke who happened to install it and found it easy to hack at a translation. I agree that supporting professional translators is good but this is not the only way to get an application translated. >> So if we have 500 templates all 500 TMX files would be loaded into >> memory? > > ditto for property files > > There are two facets of the i18n process. One is internal, where each > language template class harvests it's own language specific content and > stores it in an internal representation, in the most efficient manner > mangageable. The other is external, where humans are doing the > translation work, most probably using tools designed specifically for > the job. I think tmx files will suit both facets. I think it comes down to us seeing the translation being done differently. > Although there may be 25 languages per template, times 500, that makes a > lot of class files but they're only loaded into memory upon access. If > the system supports 25 languages but only one is ever used then only one > set of resources is ever loaded. But if all the languages are in one file then all the languages will get loaded for each template. If we used ResourceBundles then only the used languages would get loaded. Isn't adding TMX also adding more complexity. ResourceBundle is reasonably bug free. Although I'm sure you'd write a good TMX implementation it wouldn't have had the ammount of testing/tweaking that ResourceBundle has. -- +--Matthew Buckett-----------------------------------------+ | VLE Developer, Learning Technologies Group | | Tel: +44 (0) 1865 283660 http://www.oucs.ox.ac.uk/ | +------------Computing Services, University of Oxford------+ |