← Back to team overview

launchpad-translators team mailing list archive

Re: Prioritize translation types by tags


On 2010-04-21 13:26, Anders Jenbo wrote:
Marking strings as fuzzy is actualy something else that is missing. It
should be done automatic when a string changes instead of just droping
the translation, this alows translators to just make the same,
corrections to the string instead of having to retranslate it. It would
also be very help full if LP was also able to show what was changed in
the English string next to the fuzzy string.

This has come up many times. We used to support gettext's fuzzy matching, and it just wasn't good enough. At the same time it complicated what was already a very complex problem domain, so we removed it.

The basic problem is that English strings don't "change." Some strings disappear, and others appear. There may or may not be similarities in ordering and content, and some of those will be more meaningful than others. People often assume intuitively that the relationship will be obvious to the software, but as a general problem it's not.

It should be possible to develop something that will usually make helpful connections between removed strings and added ones. But we are only a few developers and it's just not something we have time for. The code is open though, so if somebody else sees a good way of doing this, I'm sure we'd be able and willing to assist them.

In fairness I should say that this is not something that you'd finish in a weekend. There's a lot of complexity that--we hope!--isn't entirely obvious from the outside. It requires changes in the UI, probably in the database, and in the code inbetween. All of those have to be carefully designed, reviewed, and tested at multiple levels. Doing a lot of work without collaboration with the developers at any stage, from specification to code review, is a recipe for failure.