Major design flaws in BonoboUI ?

[Michael Meeks <michael ximian com>]

Hi Maciej,

        Watch the Meeks get super defensive about his pet code :-)

On 20 Feb 2001, Maciej Stachowiak wrote:
> Jody Goldberg <jgoldberg home com> writes:
> > On Tue, Feb 20, 2001 at 04:06:59AM -0800, George wrote:
> > In an ideal world the bonobo-ui specification and verb/layout
> > split would be use in gnome-libs.
>
> I think the Bonobo system still has major design flaws and many
> implementation problems. We are still finding serious bugs in it

        Really !? interesting that you arn't reporting them to me, or
sending me patches. In the last weeks the only significant fix to the code
was of a minor feature I added recently to accelerate Nautilus.

> and the lack of so much as a DTD for the UI XML format makes it a very
> error-prone API for developers.

        If the lack of a DTD really bothers you, then please write one. I
personaly don't loose sleep about it, the format is documented by Havoc
and myself in bonobo/doc/ui-xml.txt for developers to use. Furthermore, it
is anticipated that GUI tools ( glade ? ) will be developed to generate
and edit the XML. The lack (?) of a DTD for a .glade file doesn't keep me
awake at night.

> I don't expect it to be ready for prime time for GNOME 2 unless some
> major peer review action of the design and implementation happens.

        Ok, well I don't understand the substance of your criticism. I
think there is much that is good in the Bonobo UI code, and there is much
that can be cleaned up in the absence of a desire to help Nautilus retain
binary & source compatibility at the cost of hundreds of wasted hours, and
shipping bonobo-1.0 with already deprecated code and interfaces.
 
        Yes, there are things that are wrong, things that are unarguably
sub-optimal, etc. etc. however, if you go and read through eg. Darin's
criticism mail, or Havoc/Owen's as of some months ago you will see that
many of the non-blue sky addressable issues have been addressed.

        Some such as 'its slow if you do 8000 menu updates where only 1 is
needed' I do not view as a bug in my code :-) Furthermore, the code has
had a fair degree of 'peer review' eg. Havoc read it through and commented
on it extensively at one stage [1]. I discussed the design with some Sun
engineers, and people at Ximian, and in the community before starting. I
am still waiting for your comments to arrive - especialy on the major
design flaws :-)
        
        In conclusion, your comments seem to be based on out of date
information, and are not comparing the Bonobo UI code with the
alternatives - none of which cut it as far as I can see.
          
        Regards,

                Michael.

[1] - Some of Havoc's initial feedback:

- much easier to do the merging than with an actual widget tree   
- extensible: we can add new XML elements or attributes later
- very friendly to GUI builders
- very friendly to user tools for customizing menus; see the MS
 Office menu/toolbar customization stuff
- very friendly to scripting languages and language bindings
- buzzword-compliant!

-- 
 mmeeks gnu org  <><, Pseudo Engineer, itinerant idiot