Re: Remove unnecessary markup from messages marked for translation

[Christian Rose <menthos menthos com>]

tis 2002-03-05 klockan 17.25 skrev Darin Adler:
> > My bug report wasn't about the ">" -> ">" bug, but nevertheless will
> > there be a new release soon with this particular bug fix?
> 
> It's in 0.17, which is already out.
> 
> By the way, the ">" -> ">" wasn't quite a bug. With this change, we can
> no longer localize strings in XML files that contain actual "<" characters
> that are not part of markup. But it's probably the right tradeoff.

By "the bug" in this case I actually meant the ">" -> "&gt;" thing.

I don't really agree about it being the right tradeoff. Yes, < and >
might be illegal(?) characters in XML. But that should only be the
application developer's concern, not the translator's.

Entity encoding worsens readability in po files, which is a problem, not
only for translators but also for the people offering to verify
translations.
But it has also another, perhaps bigger drawback in po files: intltool
does no longer hide the underlying format in the po file. This means
that in for example Gnumeric I have to translate the exact same messages
twice: in one case the message "Add ->" is stored in a C file, in the
other case "Add ->" is stored in a glade file, and so on for several
other messages. In each case, only the encoding is different, but they
are all treated as different messages and need their own "translation"
which is just an unnecessary translator burden.

Also, in the case that the developer moves a message from a C file and
rewrites his application interface partly or fully in glade (not quite
uncommon), although the messages are exactly the same, now all the
translations need to be updated.

Really, I see no benefit in the current behavior. The benefit is if
intltool is transparent with regards to the underlying format when it
can be.

What I propose is the following:

Used in glade original 	->	po file	->	Rewritten glade
-------------------------------------------------------
< >			->	< >	->	&lt; &gt;
&lt; &gt;		->	< >	->	&lt; &gt;


Opinions? Is this something that could be implemented? I'll create a bug
report about this if you want me to.


Christian