Re: Rare slashdot insight

[Miguel de Icaza <miguel ximian com>]

> This would seem to correlate somewhat with the gal situation -- discuss
> amongst yourselves. 

The advantage to develop widgets outside of the "core" of the
operating system is that you can see them mature, fix the bugs,
improve the interface and polish it before you are ready to deploy in
a situation where you will end up maintaining the code for ages.

Lets take Combo boxes for example.  A lot of widgets went into the
early Gtks without having being used in a wide range of scenarios,
which lead to the creation of forks of them.  The Gal Combo box was
born from an expanded set of needs than those provided by Gtk.

The first shot at the Gal combo box was more flexible than the Gtk+'s
one, but it also fell short because when it was written, it was
targetting a set of problems and it fell short of addressing a very
similar set of problems that could have been fixed trivially (namely
that people wanted to be able to customize all the elements of the
Combo Box, basically removing *one* pre-computed widget and allowing
it to be dynamically specified). 

A second iteration over the ComboBoxes could fix the issue (I am not
sure if today's Gal actually incorporated the change, my memory is
fuzzy). 

This is just an example of how a widget evolved.   Between the first
generation "Gal" widget to the one that was actually useful enough to
be used by multiple applications it took about six months.  And if
the new changes went it, it probably took a year just to figure out
the existance of those needs.

Miguel.