Re: dpkg i18n and gettext
Ian Jackson wrote:
> > > I'm rather unimpressed with the current state of dpkg's build system,
> > > including the i18n stuff. I'm also unimpressed by GNU gettext (I
> > > agree with gettext's detractors in favour of catgets).
> > *sigh*
> ? You don't like reorganisation ?
Well, not really. I don't like it especially when working on foreign
code. Things should be left as they are. No files should be moved
or renamed, no indention should be changed - as long as this is feasable.
As upstream author of several packages it's always a pain including
and reviewing patches that change indention and the like.
As author it is however needed from time to time to change things and
move files around.
> > > Over the next weeks and months I expect to be un-reorganising dpkg's
> > > build system to look more like the way it used to. As part of this
> > > I'm considering dropping gettext in favour of catgets.
> > Before you continue could you draw a small picture how this works.
> > a) Are there still regular strings in the source code?
> Um, I was expecting people on this list to have views about gettext
> vs. catgets, not to ask me about it ! Changing the i18n mechanism is
*sigh* Not me, I only learned gettext last weekend. :)
> about the last of the things I want to change (after stripping out
> automake and probably libtool too). I was hoping for feedback, not
> questions :-).
Hmm, does my "Please leave it to gettext" count as feedback?
> However, as I understand it (and I haven't investigated it deeply
> > b) How are translated strings handled in the source archive?
> (b) There would be a macro which would invoke catgets. The macro
> would take two arguments, instead of gettext's macro's one: a message
> number, and a default string.
So the author or maintainer has to know about the message number while
writing the code?
So the translating team needs to know about the message number when
So the translation team needs to find out when the default text for
a particular message number changes in order not to re-use old - and
now wrong - translations?
For me it looks like catges is more convenient for the author who
translates the program into all languages while gettext is more
convenient for authors who don't want to care about translations
and translation teams that are different from the author.
If so: Please leave it to gettext.
A mathematician is a machine for converting coffee into theorems.