Updating translation workflow

[Warren Block]

We have some problems with our translation workflow, and updating it 
could make life easier for everyone.  Note that as an American, I barely 
speak English, so there may be misconceptions in the following.  Please 
correct me if necessary.

* Translators work too hard.  No automation, no assistance from the
   computer to see what needs to be translated, no reuse of
   previously-translated terms.  Because of this, updating translations
   after the English version changes can take a long time.

* Few new translators are willing to work that hard, so we are missing
   up-to-date translations to several important world languages.

* Separating whitespace and content patches makes it more difficult for
   writers to edit existing English documents.

The current workflow, such as it is, is very, very old.  Since it was 
created, new ways to translate have come about.  PC-BSD is using Pootle 
(textproc/pootle, http://pootle.translatehouse.org/) for their 
documentation: http://pootle.pcbsd.org/

I think we can use a combination of tools to work with our existing 
documents.  The workflow would be like this:

1. Use textproc/itstool to create a "strings to be translated" file
    (.pot) from an XML file.

2. Use pootle as a database to translate any of those strings that have
    not already been translated.

3. Use itstool to merge the translated strings back into the XML file.

4. Build the translated document (DocBook or XHTML) as normal.

The first difficulty I've found is that itstool does not like entities. 
It could be modified to accept them, but probably the more correct 
procedure for translations is to expand them before using itstool.  This 
would be more correct for translations, as it's been explained to me, 
because some languages modify nouns depending on how they are used.  (In 
other words, our use of entities can be English-centric.)

Is there an easy way to expand entities?