Hi Jean- Christophe,
in the Q&A session you may find the answer to your question already:
http://wiki.services.openoffice.org/wiki/Translation_for_2.3#Q_.26_A
Also, please see my comments inline:
Jean-Christophe Helary wrote:
I realized a few days ago that the .sdf (at least for the fr project)
for the coming 2.3 contains weird stuff without much of an
explaination as to how to differenciate the different parts.
1) in some places the target part is made of what would be a "fuzzy"
in PO, but without specific notification of the fuzzy character
what you see is the previous translation. This means that in the
meanwhile the English text has been updated and since in most cases the
old translation contains terminology which can be reused for updating
the string, we decided to keep as a sort of *suggestion* the previous
translation instead of overwriting it with the English text.
2) in some places it seemingly contains exact matches
sometimes the English text has been updated in such a way that this is
not translation relevant. For example a typo in the English text has
been corrected. Since the authors may not necessarily know if a change
is translation relevant or not, they flag the English updated text has
updated and it gets extracted as *changed* strings when we prepare the
files to send to translation.
3) in some other places it contains the source English string
when the English text is completely new. This means that this is the
first time the strings gets translated.
In the case where the fuzzy is present, the reference links are
sometimes totally different. Which means that besides for the actual
editing of the translation, it is also necessary to edit the links.
Yes, in this case the translation needs to be updated including links,
tags, variables etc....
I wonder about the utility of such a mechanism especially since there
is no way to differenciate between the 3 patterns in the .sdf itself.
The utility is that in may cases the previous translation contain
terminology that can be reused to update the text....
It seems to me it would have been faster to _not_ insert fuzzies at
all and to provide a complete TMX of the existing OOo contents instead.
They are not fuzzies.....
Right now, if one wants to create a TMX out of the .sdf files (either
with the Translate toolkit or with Hearstome translation suite, I
suppose there are other ways though), it is impossible to have the
source strings corresponding to the fuzzy target and thus the
matching in a TMX suppotirting CAT tool will not be of much use.
You cannot create TMX out of the sdf files provided because the
translated strings contained in it are not final translations....
Is there still a way to get SUN to provide the l10n teams with TMX of
the existing contents, similarly to what we can get through the
SunGloss system ?
We could provide you with an sdf files containing the final translations
if that helps....
Rafaella
(FYI, the NetBeans team is provided with TMX and that greatly
enhances the localization process.)
Jean-Christophe Helary
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
