Re: Adjusting the 2.4 schedule

[Owen Taylor <otaylor redhat com>]

On Tue, 2003-06-24 at 17:57, Soeren Sandmann wrote:
> Owen Taylor <otaylor redhat com> writes:
> 
> >  - GtkToolbar
> 
> >From my point of view, there is no more to do on this (the toolbarmy
> earlier mail to this list), at least not as part of libegg. I can
> commit if you want me to.

Hmm, I probably should at least look over the API changes since
I last reviewed the new toolbar code first. I'll see what I can
get done on that by Friday, and if you don't here anything more
from me, you can assume it's OK to commit.

> >  - Smooth scrolling", like in Internet Explorer (#103811) 
> >    [ my impression is that more experimentation is needed
> >      here to figure out the final API ]
> 
> The new API in the smooth scrolling patch is not intrinsic to the
> smooth scrolling. It's just convenience "semantic" functions that
> change the adjustment in steps like "wheel up" or "step down" instead
> of using numbers to change the adjustment.
>
> If you're thinking of the "intention to scroll" API that Federico
> talked about on GUADEC, then note that while such API would improve
> both appearance and speed of the scrolling, it is not necessary for it
> to work.

I think that the two API's do need to be considered together - 
with an "intention to scroll" API present you need notification 
of the beginning and end of the scrolling sequence. Your API proposal
wouldn't allow this, as far as I can see, since smooth scrolling
is completely inside GtkAdjustment.

> However, if the smooth scrolling is punted to a later version, then I
> guess I'm fine with that. The patch still needs some work, and it is
> not clear that it can be unconditionally turned on (widgets
> could possibly get confused when they see intermediate states they
> don't expect).
> 
> I do think it is important that the part of the patch that changes the
> wheel step from half a page to something more sensible goes in for
> 2.4. Scroll wheels are very difficult to use currently.

Hmm, I really don't believe that the formula you use can be quite
right ... at least for something like GtkScale where the adjustment
isn't necessarily in pixels.

For proper scale invariance, the scroll step has to have the same units
as the page_size, and pow(page_size,2/3) doesn't. E.g., if the page size
is a million, then the scroll wheel will step by 1% of a page.

Perhaps the right formula involves

 page_size * function(step_size/page_size)

? Also, there is the consideration that Arnaud brought up about 
user-configurable scroll-wheel distances on Windows that should
be investigated.

Regards,
					Owen