Re: English translations stats

[Christian Rose <menthos gnome org>]

tis 2004-02-24 klockan 21.25 skrev Abel Cheung:
> Sorry if I expressed my meaning in the wrong way, what I want to say is:
> declaring England English as unsupported language just because of its
> low stat is a bit unfair, since very large portion of the strings are
> already shown in correct GB english.

We have one (1) method for counting supportedness. It is based on the
very basic task to, with the use of tools like intltool-update and
msgfmt, calculate the total number of messages that have been translated
and verified by a translator to be a valid translation in all the
affected modules, i.e. the number of non-fuzzy messages without empty
msgstrs. This number is recalculated when it's needed.

When a translator leaves a msgstr empty on purpose, there's no way of
knowing that it was on purpose afterwards. It could also be a completely
new message that the translator never has verified to "not need any
translation". The same is true for complete po files if they are left
untranslated on purpose. It's very difficult to know for a human not
involved in that team, and it's damn right impossible for a tool.

Why is it impossible? Because very simple and basic information is
lacking. When a translator fills in the msgstr of a message and marks it
not fuzzy, he or she adds the information that this is a valid
translation, and it has been validated by a human translator.

Which leads us in to the time aspect. Applications change over time, and
the translations need to do the same. Say that we took the unmodified
Swedish GNOME 2.0 translations and injected them into the GNOME 2.6
sources. After a couple of runs of intltool-update we would know exactly
how many translations were still valid. Also, we would know exactly how
many messages would need to be translated and validated.

How do you do that with translations that have been left empty on
purpose? Hint, you can't, since the information on how many messages
would need translation and validation to be correct cannot be computed
this way.

So how do we know the current level of supportedness for the en_GB
translations? Are they unsupported, partially supported, or supported?
Do we, as an example, know how the supportedness of en_GB has changed
since GNOME 2.0? New messages have come, and we don't know if they've
been left untranslated on purpose or just not taken care of. The answer
is: We don't know.

So if you claim you know en_GB is supported, I prefer seeing some hard
data on it. You can send the calculations to me. No estimates, hard
data.

And if you claim you know a better way of calculating the supportedness
of translations, a better way than what we already have, and a way that
in addition should support calculating the level of supportedness even
for translations that have previously been left empty on purpose, then
please present it. I'd bet it be a slightly problematic task, since it
would basically involve at least a computerized en_GB translator going
over all of the strings in runtime. An AI.

Until that better method is presented, I think we should stick to the
current way of calculating translation status and supportedness, not at
least because the en_GB team has realized these problems and are now
producing translations that will be possible to count and verify like
any other team's.


Good luck!

Christian