This Weekend I got a couple of Emails from Donna about the German Translation breaking (her) English original. Understandably she was quite furious. I think she and Bart might have fixed the problem for now, but behind this is an issue of greater magnitude and cause of constant frustration. Simply put the PO/POT files should help keeping the translation synchronous. For longer DocBook documents they tend to become an obstacle however -- often hindering translation altogether -- for the following reasons: - sometimes the structures of the XML files don't *exactly* match. As Bart pointed out and as I experienced myself this makes running split2po impossible and fixing it is a major timesink. Apart from that fixing this problem by keeping documents structurally the same is in some way like trying to impose the syntax of one language on another. A bit like if I translated the previous sentence from German literally. Sometimes different structure of a document would make i18n and l10n easier. - what Donna experienced is constantly happening to translators, untranslated paragraphs appear (which is actually fine, since it is part of the syncing thing and ensures that revisions of the original document are carried over in the translations), or translated paragraphs disappear (which is always a bit problematic, even if it would be correct, since following paragraphs may depend syntactically on previous paragraphs now missing, even if that is not the case in the untranslated version it may very well be so in the translation). Worse there seems to be some POT/PO magic going on that causes some mystery changes, very much of the type Donna experienced, making her think that German translators are (which in fact we aren't) adding strange German information to her English XML file. - a Skolelinux/Debian-edu peculiarity is what we experienced with the IKT-bok translation. Either the makefiles or the SGML/XML Docbook were broken, but we never got a chance to see translation results. (I think there is still no German HTML version of the IKT-bok online including all strings that have actually been translated -- please correct me if I am wrong). Also whilst translating, looking at PO files (or Docbook files for that matter, although I personally feel more comfortable about them) to proofread and for context is a bit like flying an aircraft blindfolded. I want regularly compile/transform to HTML to see what I have done and to enable me to send drafts to people who ask me for them. Our current not in all respects good solution is that we took the then current version (1.73) of the newdriftbok.en.sgml and started translating to German. This was easy enough as we are familiar with Docbook, Charsets, and have parsing editors that master both without crashing all the time (contrary to my experience with KBabel). Also we have Saxon and FOP in place to regularly produce HTML and PDF previews using the docbook xslt stylesheets. Bart and I are in contact and we agreed on keeping the German document structure as exactly same to the English as possible (living with oddities in the German translation like empty paras, etc.). So we can do split2po and use po to keep the document up to date after the initial translation is done. For the time being this seems still the best solution to me, even though I see the pitfalls (Bart: "timesink" -- Donna: "[...] to find that someone has translated at least one paragraph of my working English document into GERMAN!"). So what is this lengthy rant? 1) An apology to all who have trouble because of the Germans and their translation -- that is mainly I think Bart and Donna (in alphabetical order). Sorry! 2) Explanation of my reasoning, so Germans seem less strange. 3) Call for suggestions how to better translate Docbook files. Kind regards David
Attachment:
signature.asc
Description: Dies ist ein digital signierter Nachrichtenteil