Re: [cgmo-webcgm] SV, CV and CSS2-like inheritance [was: Re[2]: [cgmo-webcgm] Style properties]

[Lofton Henderson <lofton@rockynet.com>]

At 10:27 AM 5/18/2005 -0400, Benoit Bezaire wrote:
>[...]
>LH> [Stroke-weight is a property of an APS.  If convert-to-absolute for
>LH> stroke-weight means that you must convert % to NVDC, this cannot be 
>done at
>LH> the APS level.
>Why can't this be done at the APS level?

Example:

BegAps obj1;
    BegApsBody;
    linewidth 10;
    line 0 0 1000 1000
    edgewidth 15;
    intstyle empty;
    rect 100 100 900 900;
EndAps;

First, stroke-weight that is already nvdc is no problem.  E.g., if DOM 
applied a stroke-weight style property of 25 to 'obj1' ... no issues there.

Suppose stroke-weight is 100%.  What is the nvdc equivalent of that, at the 
BegAps element?  What it means inside the APS, at the level of individual 
line and fill primitives, is lw 10 and ew 15.  What can you say is the 
equivalent inheritable (single) nvdc value of CV of stroke-weight at the 
BegAps element?  (To make that last question more interesting, consider 
sub-trees of nested APSs within 'obj1'.)

Suppose DOM sets stroke-width to 50% on obj1.  What is the nvdc equivalent 
of that, at the BegAps element?  Etc...

-Lofton.


>LH>   So the CV of stroke-weight is ill-defined at the APS level,
>LH> if the CV requires convert-to-absolute.]
>
>LH> B.) CSS2 says that Computed Values are inherited.  (CSS3 allows each
>LH> property to define which type of value is inherited.)
>
>LH> Condition A and condition B taken together makes it problematic to apply
>LH> the CSS2 inheritance model exactly to our style properties inheritance.
>Why is it problematic?

You cannot have both condition A, plus condition B, plus keep close 
alignment with the 3 steps (or 4, or 2) of the CSS2 inheritance model.

-Lofton.


>
>LH> If we relax condition B, i.e., if we say that SV is inherited down APS
>LH> branches (for style properties) instead of CV, as allowed by CSS3, then
>LH> that inheritance model works.  (But it leaves CV pretty much unused in the
>LH> model, and doesn't solve the problem of uncomputable CV at the APS level.)
>
>LH> If we relax condition A -- i.e., convert-to-absolute happens after CV,
>LH> e.g., at the Used Value stage -- then CV is the same as SV (except for
>LH> resolution of the values 'inherit', which we don't allow anyway for style
>LH> properties.)  And inheritance down APS branches works, CSS2-like.
>
>LH> -Lofton.