[Interest] Qt 5 and filesystem

[André Pönitz]

On Fri, Jan 04, 2013 at 10:09:50AM +0100, Bo Thorsen wrote:
> Hi Stéphane,
> 
> Yes exactly, the problem is that QML is young. I expect that the
> capabilities will expand quite a lot over the coming years, and
> file handling is an obvious place for improvements. Qt (the C++
> parts) had over 16 years to develop into the current system, QML
> is an infant compared to it.

The main problem of QML developent is that there was (and to a large
degree still is) not even an attempt to re-use as much as possible of
the existing and field-tested solutions that already exist _in Qt_,
or to massage Qt into a shape where such re-use is easily possible.

I understand it's fun to do something completely new, and in times of
seemingly unlimited resources it even might look attractive. That does
not necessarily make it a good idea.

There are no unlimited resources. A "tabula rasa" approach ignoring
past problems and their existing solutions is not sustainable. Worse,
the current approach already failed to deliver on some core promises,
like "easy toolability", suggesting that the road to a full solution
will be as bumpy as any of the alternatives. The QML world will run
into a similar set of backend problems as the C++ world did during the
last two decades, and it will take several years to mature on its own.
The only viable shortcut is a conscious attempt to share solutions
between the two stacks, i.e. to have high-level C++ interfaces to
"most of Qt" accessible from _thin_ gui wrappers on top.

Qt had a tradition of finding reasonable levels of abstraction, and it
seems pretty much alive in the backend. I see no particular reason why
this could't be extended to the frontend again in 2013.  (Happy New
Year, by the way...)

Andre' 

I am speaking purely for myself.