Sahana 0.6.2.2 language pack update

[Dominic]

Three days after the stable release, I have now packaged an update of
the language pack for Sahana-0.6.2.2, easily available as usual via:

* http://pub.nursix.org/sahana/locale

Due to the changes between RC2 and the stable release, the translation
templates have also changed. Hence, most (formerly complete or almost
complete) translations have new small gaps that need to be closed.

The next language pack update is scheduled for April 11, 2009.

[Chamindra de Silva]

Thanks Dominic. Is doing a new package like doing it from scratch or
is it a small delta of effort as I can imagine not much has changed
since the last version on the 0.6 branch? If it is a lot of work, then
we need to see how we can help you make this easier.

chamindra de silva
http://chamindra.googlepages.com

[Dominic]

Hello Chamindra,

doing the language pack is in fact a lot of work - not due to many
changes in the translations, but for merging all incoming sources. The
overall process is something like this:

1. Building the PO template files

- (re-)build the sahana.pot translation template from the source
code as necessary
- split the PO template file into sections, for use at the Pootle
server
- check the template file for syntax errors

2. Checking and transformation of incoming translations from various
sources

- incoming translations may be in PO format, but may even also be in
other formats (e.g. spreadsheets, CSV, or the web2py dict format),
which need to be transformed into PO first
- incoming translations may be any arbitrary portion of the
sahana.pot file, as long as they overlap the sahana.pot file at least
at one single string
- incoming translations have different priorities depending on where
they come from (with the Pootle site as #1 source)
- incoming translations may contain syntax errors, duplicates or
other mistakes, which have to be fixed before merging in - mostly
manually
- check for specific issues that have been discussed on the mailing
lists
- it is important to keep track of all sources, and translators (who
has done what?)

3. Merging all incoming translations

- once all sources are transformed and clean from errors, they can
be merged (cascade merge), which is an automated process (I have a
couple of semi-automatic shell scripts for that)
- if the template has changed (because Sahana source code may have
changed), you need to merge old translations into new the new template
(to maximize the re-use of prior work)
- documentation of any changes, especially _when_ source code or
templates have changed

4. Building the language pack

- this is an automated process, duration approx. 3 sec on my
machines

5. Testing the language pack

- requires installation on a demo of the current release (which has
to be updated if the sources have changed)
- try out every language and browse through all modules to find
major problems (I can't read every language and script, but I can see
major issues such as overflowing boxes, wrong escape sequences)
- check for all locales and scripts (if they can be installed and
used) to find and document possible installation/configuration
problems at the server
- test for different browsers (if all characters are shown, if the
script fits into the UI etc.)

6. Publishing of the language pack

- Mainly upload to sourceforge, but also some documentation as
necessary

7. Update of the Pootle translation projects

- This takes a few minutes to update them, but sometimes hours to
save pending suggestions

8. Update notification

------------

So - there is some work in there. I always try to have semi-automatic
scripts for all tasks, especially when dealing with a lot of
repititive work (i.e., same basic process repeated for each language).

Kind regards,
Dominic