Re: [9fans] Evolving rio / GUI development

["Fco. J. Ballesteros" <nemo@lsub.org>]

> Plan 9 has a strong intrinsic
> opposition to change; this is mostly good beacuse of obvious
> reasons. But some experimentation would'nt hurt. I just can't see the

Then just go ahead and experiment!

These days we run what we call Plan B. That's an experiment
on Plan 9, the system behaves differently and we're learning
what we like of the system, and what we do not. In particular,
what we like is that the system mounts resources automatically
and that it makes it easy to adapt to changes. What we don't like
is mostly that during the auth stage, a missbehaving fileserver may
lock your system.

Before this, we experimented with redirfs, which was a previous attempt
to do this at user-level, and it didn't quite work, because not all files
were automatically mounted and failover could fail :-), doing this in the
kernel has simplified things a lot.

Regarding the UIs,
we have what we call omero (another experiment), it's a mixture of
a window-toolkit and a file server, that handles all the graphics by
its own. The application creates files to build its UI, and the fs takes
care of the mouse and everything else. This allows you to replicate
and move around UI components without placing the burden in the
application.

During the construction of omero we tried two nice experiments, one
was mouse cooked mode, which I now consider as a mistake, after
experimenting with it. The other, which I consider a big success, is
defining the left control, start-button, left-alt as mouse 1, 2, and 3 buttons.
We use the menu for composing (old alt), and capslock as control.
This made it very nice to use 3-button interfaces (like acme and omero)
in laptops.

Probably sometime this year we'll consider both the system and omero
stable enough to try and talk about how to bring this stuff into Plan 9, if
anyone cares. Even if others consider that it's not worth placing this
kind of stuff in Plan 9, some things might still be useful, like the mouse
buttons thing.

My point is that experimentation is a good thing, so go ahead. Of course
it's perfectly nice for others to say `we won't use the resulting code', but
that's why it's an experiment, isnt it? It may also be that those that
refuse to use the resulting code might teach you a better way of doing it,
which is even better.

good luck

PS: Sorry about the long mail :-)