FESCo Meeting Summary for 20090424

[Matthew Garrett]

On Thu, Apr 30, 2009 at 07:16:38PM -0700, Adam Williamson wrote:
> On Fri, 2009-05-01 at 02:56 +0100, Matthew Garrett wrote:
> > Does it? The desktop team would presumably assert that the cost of the 
> > reduced user experience and increased user confusion is sufficient to 
> > make the product *worse* than it would without shipping two mixers. 
> > There's no absolute truth here, so the question isn't which option is 
> > worse. The question is who we trust to make that decision.
> 
> Interestingly, I haven't seen anyone flat out make that assertion.
> 
> I don't see how it could possibly stand up, either. I mean, the
> alternative they're arguing for is to stick with the current state: a
> release note advising people to use alsamixer. So instead of having two
> graphical interfaces, they're apparently happier to have one graphical
> interface and one obscure and extremely user-unfriendly curses
> interface. I really can't quite grok how that's better.

The basic counterargument is that providing two mixers in the menu is a 
cost to every user of Fedora. This cost is being justified on the basis 
that the cost to some unknown number (but assumed to be a minority) of 
Fedora users of *not* having that mixer is large. We have no numbers to 
attach to any of these costs or the proportion of the user base 
affected.

The people in the best position to make judgements about those costs are 
the desktop team. They've got more experience in this field than you, me 
or pretty much anyone else on this list. If they make an assertion about 
these weightings then I'm not going to feel in a position to argue 
against that until I've got some numbers, just as I'd be inclined to 
ignore them completely if they started contradicting me on power 
management issues without providing some numbers. Why do you feel 
differently?

> Believe me, alsamixer is user-unfriendly. I have several years of
> experience of advising people to use it. I started out saying "oh, just
> run alsamixer and do XYZ..." After aforesaid years of bitter experience,
> I now say something like "run alsamixer. You might need to do alsamixer
> -c0 or -c1 or -c2 or -c3, it depends how many cards you have. You can
> scroll to the right by pressing the right arrow a lot of times to see
> more controls, even though there's no scrollbar indicating this. You
> toggle inputs on and off by pressing space, but you flip switches by
> pressing m. To get to the capture inputs you either run alsamixer with
> -Vcapture, or hit tab after running it."
> 
> I'm just at a loss to try and grok how this is considered preferable to
> an alternative graphical mixer. Where's the win?

I don't think anyone's actually argued against there being a separate 
graphical mixer. I think they've argued fairly strongly against it being 
in the menu by default. But I think you're asking the wrong question. 
The right question is "Who is going to need to run this program", and 
the followup question is "And what kind of interface do *those people* 
need?". If a large number of naive users need to run gst-mixer then 
we've already failed badly. Adding it to a menu doesn't fix that. If the 
users who need this functionality tend towards being advanced then the 
nature of the UI is less important than the functionality is exposes.

> > Having said that, there are ways in which the Ubuntu decision is a 
> > better one than ours. It's less flying-car future, but there's no 
> > regression and no additional user confusion. I've personally got no 
> > objection to FESCO overruling a decision on technical grounds and 
> > forcing a reversion to the previous status quo (or whatever else is 
> > available under the contingency plans), but the compromise we've ended 
> > up with here looks awfully like a baby that's been cut in half.
> 
> I don't really think so. We all seem to agree that the new mixer
> interface is The Future and should suffice for most people, but not for
> a significant minority. We want people using the new mixer interface to
> make sure it's ready ASAP. So making it the default that most people
> will encounter and be happy with, but having a relatively
> easy-to-discover fallback for those whose needs aren't covered, seems
> sensible to me. Given the choice between reverting to the old g-v-c by
> default or this compromise, I'd probably choose the compromise.

How is the user supposed to work out which is the fallback? Intuit it 
from the fact that only one of the two seemingly equivalent choices in 
the menu is started from the panel applet? How does the user work out 
what the interaction between these two applications is? How do you 
prevent the users who have a working pulseaudio setup from finding 
gst-mixer and then breaking their systems? How do they recover once 
they've done so?

My somewhat arrogant assertion here is that unless you've been asking 
that sort of question then you don't understand all the issues involved 
in this decision. And that implies that your opinion on the matter is 
about as relevant as my opinion on QA workflows.

That doesn't mean that you have no input on the matter. Providing well 
documented descriptions of failures or inadequacies in PA is helpful in 
ensuring that it ends up as a better project that fills as many of the 
needs of Fedora users as possible, but at some point you have to trust 
that software maintainers know what they're doing. And if what you've 
ended up with is something that the project as a whole doesn't find 
useful, then the best thing to do is to avoid using it rather than 
forcing UI changes upon an unwilling designer.

-- 
Matthew Garrett | mjg59 at srcf.ucam.org