XComposite input redirection/transformation proposal

[Rémi Cardona]

Kristian Høgsberg a écrit :
> I don't really think it's that bad - a 3D tri-mesh mirrors the way we
> do output redirection, that is OpenGL triangles.  I'm sceptical that
> we'll need more advanced primitives for the foreseeable future, I just
> don't see a use case.

Neither do most of us, because experimentations in that area have been 
very limited. Here's how I see it :

  - going down the server path means "forcing" a certain type of 
primitives, that may very well be enough for the next 20 years, but OTOH 
could prove to be very limited for other unforeseen cases (ie, non 
linear transforms such as zooms). Should that be the case, we would have 
to add new primitives to the extension, or even deprecate the extension 
altogether, yet maintaining it for backwards compatibility.

  - going down the client side is more complicated, everyone seems to 
agree on that. Yet, it allows clients to decide whatever they want to do 
with input events, freeing the server of complicated code paths. Also, 
I'm not really fond of having to tell the server what my windows look 
like, in a way that may not match on a 1:1 basis with my OpenGL objects.

Keith and Barton [1] have already acknowledged that doing fonts in the 
server was a bad idea and that client-side was better. I'm afraid 
server-side input-redirection will be a similarly bad idea.

[1] during the recent Google Tech Talk

Cheers,

Rémi