Re: Comments on Artlib backend

[Richard Frith-Macdonald]

On Wednesday, August 28, 2002, at 06:13 PM, Alexander Malmberg wrote:

> > > > > There is an implementation of DPSarct in ARTContext.m, could this be
> > > > > used as the starting point to implement this method in GSGState?
> > > >
> > > > It's implemented using DPSlineto and DPSarc, so it could be used as 
> > > > is.
> > > > I'll move it to GSGState.
> > >
> > > Actually, since it's implemented this way, it could be moved all the 
> > > way
> > > to NSGraphicsContext in -gui. Should I move it there?
> >
> > I'm not authorative, but in my opinion it should NOT be moved to the 
> > gui.
> > In my view it belongs to gsc. See my other mail. (The gslib backend has
> > it's own implementation of this method.)
>
> Nothing prevents a backend from overriding the method to provide an
> optimized version (in fact, I already do this in back-art for some gsc
> methods). The NSGraphicsContext implementation would only be a default
> fallback implementation for backends that don't want to implement it
> themselves.
>
> In this case, it could be done faster in GSGState, but I don't think
> performance of arct will be an issue, so I see no reason not to put it
> in NSGraphicsContext.

I agree, that's the correct principle ...

We should put as much as possible into the gui rather than the backend
(to minimise the work needed to write a backend), but make it easy for
the backend to override stuff with optimised implementations.

That way, people can port quickly and experiment with new backends 
quickly,
but we don't compromise the performance of any backend.

So the issue should not be whether or not to implement stuff in the gui
library, but how to implement it in the gui in such a way that it can be
used by *or* replaced by backends efficiently and easily.