[cairo] surfaces and backends

[Mike Emmel]

On 5/21/05, Owen Taylor <otaylor at redhat.com> wrote:
> On Sat, 2005-05-21 at 11:08 -0400, Mike Emmel wrote:
> > Look like your trying to sneak the g type system and GObject in.
> > Why don't you just do a simple api that could map onto GObject
> > if someone wanted ?
> > It would be nice if the Cairo implemention could be GObject based
> > if wanted. With some work this could be a compile time option
> 
> I think my proposal is about 10 lines of code to implement, total.
> While it shares a bit of superficial resemblance to GType its
> not in any way comparable.
> 
> My suggestion actually *simpler* to implement than Carl's suggestion.

I did not disagree with that its the right thing IMHO.
> 
> In terms of making GObject optionally based on GObject, I can't
> see how that would work. The code would be different internally
> and the API would be different.
> 
> Regards,
>                                                 Owen
> 

Not sure about that I've been thinking about how to gobjectfiy not
gobject based code. I agree that it would be different if it was
gobject based i.e you would use the gobject conventions. But as long
as there is the concept of type and new and thats what exposed the
fact that your gobject internally is prob okay. Basically when your in
the situation that you have gobject its really nice to go ahead and
use it.

I'm thinking that the base requirment to back code with gobjects is
1.) Must construct via a function.
2.) Must have ref counting  functions.
3.) Type system via macro call ?
????


I think I let my internal musings about the situation leak out when
you made your suggestion :)

Mike

> 
> 
> BodyID:84520257.2.n.logpart (stored separately)
> 
>