Re: Make GNUN more user-friendly

[Thérèse Godefroy]

Le mercredi 28 août 2013 à 11:01 +0000, Ineiev a écrit :

>  > One of the problems is the huge number of useless files in the local www
>  > directory -- useless for a translator, that is. Selective checkout may
>  > be part of the answer
> 
> Yes. I think it would be useful to finalize it in a script or a page
> of documentation.

I do have a small script which does that. The only problem is, when
there is a new French translation, the po, pot and html have to be
checked out separately so that cvs can update them. No big deal, tho.
> 
>  > Another problem is the awfully complicated page generation system. When
>  > a translated page is regenerated locally, the POT is often regenerated
>  > as well,
> 
> This sounds like a bug. do you know how to reproduce it?

This happened most of (all?) the time before I started restoring www
after a validation, quite a few months ago. I could try again after
reinstalling what's needed to use the "big" GNUmakefile.
[...]

> Ideally, validation errors should be fixed in www;
[...]

I don't quite agree. Correcting errors after GNUN reports them on
trans-coord-discuss (if this is what you mean) puts a useless load on
the shoulders of this poor server.

Anyway, I find it easier to modify POs and correct the errors in www-fr
rather than www. There are less merge conflicts if another version comes
up while I am updating a PO to the previous one (possibly due to Git
being more tolerant than CVS?)

> It looks like there are some trade-offs that may want reconsidering;
> apart from what I've already mentioned, I'm not absolutely fond of its
> interactivity --- as for me, it reduces the flexibility,
[...]

I agree that the configuration should be done beforehand. It shouldn't
be part of the script.
While we are at it, why not make a nice-looking deb out of all that, and
use Debconf? Too bad for the other distribs.  ;)

For the validation itself, the interactivity usually amounts to clicking
an icon in my dashboard, then pushing "ctr-x" after checking in nano
that the list is correct, then choosing "1" for the commit message, all
within a few seconds.
If there is only one PO, I enter it in the command line, then choose
anything but 1, 2 and 3 and write the commit message as usual.

[...]
> I must note that it's unbelievably impressive for a newbie.

A newbie at Bash, certainly. Now, I did write very small programs in
Fortran for an IBM 18OO (with punched tape and a compiler the size of a
washing machine), and had to use an old Wang 720 (an oversized hand
calculator, kind of) in the late 70's because my boss wouldn't buy an
Apple II (also had some fun with the Commodore 64, etc.) Sounds ancient,
doesn't it ?  :)

Thérèse