Difficult to implement CSS1 bits

There are a couple of bits which appear,  at least on first glance,  to
either be technically difficult to implement or in some cases,  at least
inefficient.  I'd like to both complain a little,  as well as look for
suggestions at to the proposed implementation of the features.

I've spent awhile reading the description of the various background
attributes in the style sheet.  What particularly concerns me is the
notion of "fixed" backgrounds.  This appears to be meaningless (or at
least pretty useless) with the exception of the toplevel of the
document,  i.e. a BODY or HTML background.  Was it really the intention
to allow fixed, non-scrolling backgrounds on things like EM?

Even for a BODY background acting as a watermark, there it is somewhat
inefficient requiring constant recompositing and redrawing for small
scroll increments.  For any UA which does not do it's rendering
offscreen (c.f. ours doesn't yet) this leads to a constant flashing as
the user attempts to scroll the display.  The effectively requires that
the UA allocate offscreen memory for it's rendering to avoid the
flashing and recomposite the page for every increment.  This increases
memory needs from 256K (400x600 1 byte per pixel) to around 1 Mb for
us with 600x800,  3 bytes per pixel.  OK,  I'll grant you don't
"require" it for core compliance.

Another area I simply don't understand very well is the notion of
carrying margins/border/padding into inline text.  I understand how it
could be used for atomic inline elements (e.g. IMG)  but elements like
text,  which can be broken across line boundaries are pretty nasty to
have this sort of capability,  which appears to be a unique capability
as compared to any other text processing package.  Perhaps someone can
enlighten me as to how this can be implemented in a simple an elegant

Doug Rand <drand@sgi.com>		(508) 567 - 2217	
Silicon Graphics/Silicon Desktop	http://reality.sgi.com/drand
Disclaimer: These are my views,  SGI's views are in 3D