Sorry about the delay here. It got lost in my big backlog.

I don't know about MIF and PCL (yet) but SVG supports the matrix()
function in the transform attribute to directly use a transformation
matrix. But since AFP doesn't seem to support it, it may really be best
to fall back to the minimum supported by all renderers which will
probably be a list of simple transforms like translate(tx, ty), scale(sx,
sy), rotate(angle, [cx,cy]). I don't think we need skew. Basically this
would be a subset of the functions defined by 7.6 "The transform
attribute" in SVG 1.1 [1]. So, the current "ctm" trait could actually
be replaced with something like a (scaled-down) SVG "transform"
attribute. That shouldn't be difficult to implement. As a nice
side-effect, this could even be reused in the upcoming intermediate
format to do imposition (rotation, n-ups and such).

I assume you will handle the implementation, right?

[1] http://www.w3.org/TR/SVG11/coords.html#TransformAttribute

On 23.12.2005 14:04:59 Manuel Mall wrote:
> On Mon, 19 Dec 2005 04:50 pm, Jeremias Maerki wrote:
> > I see absolutely no problem with creating that special class that
> > holds all the information about the setup for the new coordinate
> > system established by a viewport. You could argue that the CTM is
> > calculated too early in the layout managers. So far all renderers
> > were pretty much ok with only working with the CTM. If that doesn't
> > apply to all renderers we have to adjust for that.
> >
> 
> What I don't quite get is why both of you talk about a new class. The 
> spec defines all the necessary traits required. What is the argument 
> against simply adding these traits to the area tree? Given that 
> Viewports are not appearing very often in the area tree (compared to 
> line, text, block, space,...) adding a few additional traits shouldn't 
> be a size/space problem.
> 
> BTW, so far we only have 3 renderers using the CTM (PDF, PS and Java2D). 
> It remains to be seen if other renderers currently in the sandbox (mif, 
> pcl, svg) can live with it or require the base properties from which 
> the CTM was calculated.
> 
> > On 18.12.2005 17:42:01 Andreas L Delmelle wrote:
> > > On Dec 18, 2005, at 14:52, Manuel Mall wrote:
> > >
> > > Hi Manuel,
> > >
> > > > <snip />
> > > > I think this needs improvement but what is the best strategy
> > > > here?
> > >
> > > I just had an --admittedly very vague-- idea about this, don't know
> > > if it is feasible. Jeremias, correct me if I'm wrong...
> > >
> > > What about creating an object type, especially meant for use with
> > > the render-type AFP, that stores the needed info? Something like a
> > > mapping from 'viewport with id'->'reference-orientation' (maybe
> > > only for orientation other than zero). That way, you wouldn't need
> > > to reverse engineer the values from the CTM. In fact, the AFP
> > > renderer could bypass the CTM completely and get the relevant data
> > > from somewhere else without having to perform recalculation.
> > >
> > > Not really sure about the how, when or where, though... As I said,
> > > very vague.
> > >
> > > Just a thought. Hope it can be of use.
> > >
> > > Cheers,
> > >
> > > Andreas
> >
> > Jeremias Maerki
> 
> Manuel



Jeremias Maerki

Reply via email to