[sugar] New approach to the controls implementation

Marco Pesenti Gritti mpg at redhat.com
Fri Apr 27 09:59:49 EDT 2007


Hello,

the idea to write our own set of controls on the base of
HippoCanvas was a bad one. I still think that would have been the best
approach from an architectural point of view. Also I think that's the
way gtk should be going on the long term. But:

* We should be leading this change upstream rather than doing our own
thing which will be incompatible with current gtk and also with gtk 3.0,
whatever direction they actually choose.

* The time is not mature for the switch. HippoCanvas miss a lot of
features that will be necessary to replace gtk widgets. We don't have
resources to add these features.

* Writing controls is *hard*. It takes a lot of time to get all the
behaviors and states right.

* We need to be able to do the controls work on demand, so that we can
keep our main focus on features.

These are the main reasons of the direction change. I'm *sorry* to not
have put more thinking into it before, I really screwed up there. I
guess the lesson to learn here is that, even if we are doing innovative
design, we should always try to do it extending our platform, rather
then replacing.

Going more in the concrete of the new approach. There are three main use
cases our graphics API will have to cover:

* Free form views (html/flash like).
* Custom controls implementation.
* Controls framework.

The controls framework will be based on gtk. We will try hard to use gtk
API and semantics where at all possible. We will extend it when it make
sense. We will patch gtk where it's possible to get the changes
upstream.

We will use HippoCanvas for custom controls implementation where it's
useful but we will not expose it in the API.

We will use HippoCanvas to implement free form views (ex. the journal
one). We can embed gtk widgets in it for the controls.

I, Eben and Tomeu had three day meeting in Milan last week to go over
the whole controls design specification and figure out the general API
requirements and framework. The implementation work already started. I'm
going to post a separate mail with more information for the activity
authors.

Marco



More information about the Sugar mailing list