Re: AWT

[Jeff Sturm <jsturm@sigma6.com>]

Thomas Reilly wrote:
>    1.Native windows can be heavyweight, so it's undesirable to have too many of them.
> 
> Doesn't see to be right.  I use this plotting tool called xmgr that
> uses tons of Motif widgets and has never seemed sluggish.  My
> impression with GTK+ apps is that they are even more snappy than Motif
> so I think we can safely say this point isn't valid.  If it is a
> problem then the place to fix it is probably in GTK+, which is
> possible thanks to free software.

Yeah.  I blame Sun marketing for dumb statements like this.  It's like
saying "lightweight windows are faster because they're lightweight." 
They obfuscate the real issues.

>    2.Native windows are opaque, so they can't be used to implement transparent regions.
> 
> This wouldn't seem to be super important.  Does anybody know if GTK+
> has support for this?  Doesn't the X11 Shape extension allow you to do
> all kinds of arbitrary clipping that could be used for transparency?

Transparent windows are useful for certain things, but don't have to be
the default.  We don't have to rely on the peer system for transparency,
the lightweight model developed in 1.1 will handle it just fine.

SHAPE is most often used for toplevel windows I believe, which are
always heavyweight (even in Swing).

>    3.Native windows are handled differently across platforms, so the AWT has to struggle to maintain a consistent view across these varied platforms.
> 
> Herein probably lies the true answer, they had a hard time porting
> everything and keeping the Windows and X11 implementations consistent.
> GTK+ pretty much rules out this one too.  And if platform consistency
> problems do crop up they can be fixed in GTK+.

Agreed.  The peer implementations in 1.0 were riddled with bugs and
inconsistencies.  Sun apparently chose to discard the peer model rather
than fix the implementation.

-- 
Jeff Sturm
jsturm@sigma6.com