Re: Reg. integration of patch for bug 109091

[Dafydd Harries <daf muse 19inch net>]

Ar 17/02/2004 am 09:49, ysgrifennodd Jody Goldberg:
> On Tue, Feb 17, 2004 at 12:50:14PM +0000, Dafydd Harries wrote:
> > To me, this does not make sense. Adding a string without marking it
> > for translation is, in my opinion, worse than adding a string and making
> > it translatable. User-visible strings which are not marked for
> > translation are bugs. Therefore, adding a non-translatable string means
> > introducing a bug.
> 
> In the general case I agree with you.  However, nothing is ever
> quite so crystal clear.  In this instance it is a bug to not warn
> the user that the data will be lost.  In star trek land our heads
> would burst into flame about now :-)
> 
> We're left judging which bug is more severe, un-translated, or
> un-warned.

Well, either:
- no message, in which case the bug is no warning
- untranslatable message, in which case the bug is that the message
  can't be translated (of course, it's questionable whether a warning
  you can't understand is any use :))
- translatable message, in which case there is no bug

I think the third case is preferable. :) The second case is least
preferable of all, not because the bug is worse but because you can
choose the third case instead and have no bug.

Of course, the latter two cases are more problematic during the string
freeze, so we have to make a decision on whether it's preferable that
the bug remains or the string freeze is broken.

> > Developers need to have a mindset where they try and get the strings to
> > a stable state earlier than the rest of the code.
> 
> I've definitely been more lax on this release than previous due to
> the slushy nature of gtk right now.  With underlying libraries still
> in flux the string freeze has come early.

Hmm, I can see how that could make things worse. If GTK+ stabilises
late, other things will be late too. I expect the same applies to other
core libraries. Does that mean that there's an extra onus on developers
of core components (i.e. ones which have many dependents) to stabilise
early?

-- 
Dafydd