[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Fedora 18 issues with translations and keymaps



On Thu, 2013-01-03 at 17:24 -0500, Chris Lumens wrote:
> > Well, yes, in the context of whether we ought to make it into a released
> > product yet. :) I didn't say it was a fair comparison. I explicitly said
> > it wasn't. But 'is it a fair comparison?' is not the operative question
> > in the context of trying to decide 'is what we have releasable?'.
> 
> This is just setting us up for stagnation and failure, though.  Such a
> comparison means we can't ever ship a significant change unless:
> 
> (1) Volunteer support appears to take care of things that we simply
> cannot do; or
> 
> (2) We wait until the last minute and when such support does not appear,
> make drastic changes to deal with the fact that there was no such
> support.
> 
> Developing this product is an iterative process.  It's never going to be
> perfect on the first go-round, and if we keep raising the criteria for
> what qualifies as good enough, we are simply never going to be able to
> release.  The feeling I'm getting here is that some previous version of
> Fedora is the gold standard and anything in anaconda that's different
> from that is cause for holding up the release.

I asked a question, I didn't suggest an answer. It seemed important to
ensure the issue was properly flagged and discussed.

> > FWIW, I actually looked at them all, and my take is that anything under
> > 40% is very likely to be pretty much unusable.
> 
> We can consider weeding those out, but you need to consider this
> question:  At this point in the release, is this more important than
> fixing partitioning bugs/NFS mounts/problems that will hit every single
> person?  If not, we should just not worry about it and consider what can
> be done for F19.

Is a fish more important than a motorcycle? I mean, it's a pretty
impossible comparison. Do we care more about native Portuguese speakers
or people who do NFS ISO installs to remote systems? How do you possibly
go about answering that question? I think the 'is X or Y more important'
stops making sense at _some_ point, and you just have to consider a
given issue on its merits.

> > we're not doing
> > weeding of 'supportable' keyboard layouts in anaconda any more, we're
> > just throwing the kitchen sink at the user, but we haven't made sure the
> > selected configuration can actually be applied properly (viz the console
> > layout problem).
> 
> Years of handling "my favorite keyboard layout isn't supported!" bugs
> means we are now just providing everyone everything.

Yes. I am perfectly happy with us providing all the xkb layouts for
people to pick from. That's what GNOME does. Fine by me. What I'm
worried about is the very problematic system of mapping xkb layouts to
console layouts for console input.

> > On the country/keymap thing, a concrete suggestion: I am not sure it's
> > even a good idea to _attempt_ such magic. How about something simpler?
> > If you pick any country other than the good ol' U.S.A., we force you
> > through the keyboard spoke as the next step, before sending you to the
> > hub. It's against the Awesome Hub/Spoke Design, but practically
> > speaking, it may be a better option.
> 
> And then potentially the network spoke, and that's three things before
> you even get to the hub, at which point why did we even bother with this
> redesign?
> 
> Right now we do have a checkbox on the welcome screen that allows for
> setting the keyboard to the default layout, whatever that is.  

That's the 'magic' I'm saying is a rather slippery thing to do. Mapping
from 'What country are you in?' and 'What language do you
speak?' (themselves two _significantly_ different questions, which we
smoosh together somewhat problematically) to a keyboard layout is a
massively complex undertaking. Take the two examples that show up all
over my 'what's busted' list: Spain and India. We have four or five
languages spoken in Spain in our language list, and four or five
associated keyboard layouts exist (in xkb, not all of them in kbd).
Someone who's in Spain might speak Spanish (Castilian), or Basque, or
Catalan. Someone in France might also speak Basque or Catalan, though.
Someone in Spain who speaks Catalan might nevertheless *type* it with a
Spanish layout, but someone in France who speaks Catalan might type it
with a Catalan layout (there is a dedicated Catalan keyboard layout, but
most Spanish Catalan-speakers don't use it). The problem in India is
much the same, only bigger: there's dozens of languages spoken in India,
many of which are also spoken elsewhere, all with different keyboard
layouts, and permutations of all factors. This is the kind of mess
anyone who tries to map countries to keyboard layouts is letting
themselves into, and why I'm not entirely convinced anyone (not just
anaconda) really wants to be doing it...anyway, that turns into a pretty
big discussion, I think.

> My hope
> is that mapping is good enough for this release.

It, well, has problems. I've filed bugs for most of them.

>   Somewhere, we have
> floating around a mockup for how to add some sort of keyboard selection
> to the welcome screen, but that is going to have to wait for F19.

Sure, I wasn't talking about F18 in this part of the mail.
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | identi.ca: adamwfedora
http://www.happyassassin.net


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]