Forum .LRN Q&A: Translation server upgraded

Collapse
Posted by Peter Marklund on
The OpenACS and .LRN translation server has been updated to run on the latest CVS HEAD sources and is now accessible at http://translate.openacs.org. There are new messages to translate, and the translation admin UI has seen a lot of improvement.
Collapse
Posted by Luigi Martini on
I can't find the expressions in their "real life": even something quite plain as the verb "is" will be translated in different ways, depending on the situation: once more I would stress this point: translating is not only a matter of words but phrases, instead. Mostly. I tried with

Files that use this message
Show | Hide files that use this message.

but could not find anything useful, only strange expressions that maybe refer to programming.
Maybe I'm just unable to navigate .LRN

Collapse
Posted by Carl Robert Blesius on
Luigi,

those strange expressions refer to files on the file system and the lines within the files where the terms appear.

Our lead translator in Heidelberg has the exact same complaint: they are not much help in finding the context (where the term appears) without knowing something about the technical backend.

We hoped to fix this in the second round of i18n work, but it seems it is a substantial engineering problem and it was one of the few things that was not fixed. Maybe Peter can provide us with some enlightenment as to why it's so hard (from an engineering perspective) as well as some suggestions so we can try to get them into future versions of acs-lang.

Collapse
Posted by Peter Marklund on
Luigi,
you are right, we still have messages that are too short. We need to do some more work merging messages together into longer phrases for better context.

Apart from that, the best way to provide context that we have come up with is the translator mode. However, translator mode has been broken by the new HTML quoting of the templating system.

Lars has an idea for a new approach to translator mode that would allow you to browse the UI and see all the messages that need translation listed at the bottom of each page. It wouldn't have inline translation links as the old mode, but it wouldn't suffer from its problems either and be more reliable.

Collapse
Posted by Luigi Martini on
I understand that there are problems, still. Nevertheless we could try and get around that if you describe to us, in easy steps, how to find those mysterious "files on the file system and the lines within the files where the terms appear."
Would it be technically faisible? I know that it would be a bit boring for the translators. It doesn't mean that we will do an in-depth browsing for every word though: in case we are in doubt, we'll have the chance to check, at least.

As far as I understand, the community believes (and hopes) that this new release will be an important step for openACS (and .LRN). So, in order to try and help, I'm not scared about browsing a bit, if you teach me (us) how to.
Also, as far as I am concerned, there is much to learn about the package, and this could be a way to explore it.
Technical issue: in case I need to have openACS installed on my computer, no way to help because installing it is far too difficult for a beginner without proper documentation. I couldn't, until now.

Collapse
Posted by Luigi Martini on
Peter, I agree that "the best way to provide context that we have come up with is the translator mode". One humble thought about it is that there are lots of pages that I cannot explore, due to my personal lack of skill, or to the fact that they are (just an example) error pages or messages, and do come out only under certain, perhaps rare, conditions.
Maybe an artificial dump of all the pages, generated with all the combinations, would be better: at the end the number of pages I have to explore should be about the same (unless I am wrong). But the good side of it is that I have a numbered list of pages, well ordered from 1 to n, and can work sequentially through the list. While at present I randomly browse the live pages, exploring here and there, returning more than once to the same page, and maybe inadvertedly skipping many (semi-)hidden pages and/or expressions.
Collapse
Posted by Matthias Melcher on
I hope the Peter's statement "the best way to provide context that we have come up with is the translator mode" is not meant to be a surrender in the struggle of finding contexts?

If the system is well designed (and I think it is) there should be some possibility to drill down the "Files that use  this message" to only .adp-files. If some other module uses a given string and transfers it via some hidden mechanism to the .adp page, this relation should be able to be put into the database.

Example: string dotlrn.user_portal_page_file_storage_title ("My Files") is found only in "dotlrn.info". In some mysterical way, it is nevertheless displayed on dotlrn-master.adp where non-programmers can see it in context. In the message catalog, there could be a pointer to some intermediate module or variable (say, navprocs.tcl or @navbar@), and there, in turn, could be pointers to other modules/variables, until eventually the final end user .adp-page turns up in the translator's dialogue.

Until this context finding problem is sufficiently solved, translators URGENTLY need help to follow these winding roads of strings. I compiled a long list of strings that are simply not understandable without their context,  http://www.rzuser.uni-heidelberg.de/~x28/i18n/blindflug.txt

Trying to resolve them is not possible even by using the source code. For example, acs-subsite.lt_Then_you_should_eithe leads to /web/translate/packages.old/acs-subsite/www/pvt/home.adp , but browsing the cvs to http://cvs.openacs.org/cvs/openacs-4/packages/acs-subsite/www/pvt/home.adp ,the string is not there.

(If the "packages.old" indicates that the message catalogue is not properly cleaned up, this should be immediately done.)