The Translation Project

I would like to propose a method of analysing or profiling string usage 
within Gettext.

The idea is simple, instead of just looking for a translated string in 
the MO file gettext will also record the fact that this string was 
requested by the application.  This data could then be used to see which 
applications are actually used by a user and what strings are indeed 
seen by the user.

Why would this be usefull?
* It allows anyone to really see what applications are used by end users 
and focus their translation efforts their.
* It allows translators to focus on key strings within an application 
first, leaving items such as rarely seen error messages for last.
* Small translation teams can quickly move their language to a usefull 
state by focusing where it counts.
* Teams can make focused reviews of highly visible parts of their 
translations.

How to implement it?
* Within gettext as a compile time option or as a runtime switch

For this to be most usefull it would need to be run on many end-user 
computers.  A translator or programmer might not be the best people to 
profile as their applciation usage might be different from an average 
desktop user.

What are the problems?
* Privacy, how do you request that a user enable profiling?  Maybe a 
configuration request when someone enables a language for their 
environment.  An option in Gnome and KDE and an environment variable for 
command line people.  Enabling it by default would probably be a privacy 
issue as you can then analyse what apps people make us of.
* How do you enable getting the data back?  Maybe leave that to internal 
- ie if someone enables data gathering then they enable getting the data 
back.  Either use email or http and encapsualte the data as an xml or 
flat text file. 
* Which version of the translations is this data refering to?  Might be 
usefull to be more pedantic about versioning translations so that you 
know what the profile data is actually refering to.
* Where do you send it? Another header option which specifies the how to 
send the data and where - via email or via http and specifies how 
frequently the data gets sent.  Maybe a cron job to parse the database 
and send the data daily.

-- 
Kind regards
Dwayne Bailey

dwayne@...   +27-21-448 7827 / 9265 (work)        
+27-21-448 9574 (fax)     +27-83-443 7114 (cell)

"An English-only, or even an English-mainly, policy necessarily condemns
most people, and thus the country as a whole, to a permanent state of
mediocrity, since people are unable to be spontaneous, creative and
self-confident if they cannot use their first language"        
         Dr Neville Alexander - 'Where English can Serve but not Empower.'

Translate.org.za - Opensource software for all South Africans
A project of the Zuza Software Foundation
Zuza - given freely, get as a gift, obtained freely

>I will break iso-3166 down to a small iso-3166.pot for just the country
>names, and a seperate one for ISO-3166-2, that is more optional, and
>submit iso-639.pot, iso-3166.pot to TP (would you also translate 
>iso-4217.pot, the currency names?)
>  
>
iso639 - please don't ignore the 3 letter language codes, quite a few of 
these are comming into use.  An option might be to split the POT into 2 
and 3 letter codes although that might make it harder to use the lists I 
suppose.

-- 
Kind regards
Dwayne Bailey

dwayne@...   +27-21-448 7827 / 9265 (work)        
+27-21-448 9574 (fax)     +27-83-443 7114 (cell)

"An English-only, or even an English-mainly, policy necessarily condemns
most people, and thus the country as a whole, to a permanent state of
mediocrity, since people are unable to be spontaneous, creative and
self-confident if they cannot use their first language"        
         Dr Neville Alexander - 'Where English can Serve but not Empower.'

Translate.org.za - Opensource software for all South Africans
A project of the Zuza Software Foundation
Zuza - given freely, get as a gift, obtained freely