Jin S. Choi writes:
> >>>>> "Henry" == Henry Baker <hbaker@netcom.com> writes:
>
> Henry> Since Postscript level 2 has garbage collection, it should
> Henry> be quite reasonable to move _most_ of the computation into
> Henry> the printer!
>
> One reason this might not be desirable is that you are then tying up a
> possibly shared resource with computation that would probably be done
> faster on your own machine.
>
> I bet my Pentium Pro and Scheme48 would beat out my HP Laserjet in ray
> tracing those metallic spheres, although it might be a close race :-)
Unfortunately, I'm not sure that the decision is quite this
simple.
Here are a few complications:
1. Some decisions can be made in the Scheme interpreter, but
some can only be made in the rendering device. For example,
different rendering devices may have different display/paper
sizes and different color/shading capabilities. It's not clear
that the scheme code *can* know this, although it should be
easily determinable on the rendering device. I have recent (and
unpleasant) experience attempting to print several long manuals
which were sized for A4 paper when all our printers had was
letter paper. I left the job to run early Friday afternoon and
went home for the weekend. On my arrival Monday morning, the sys
admins greeted me with word that not only was my job incomplete,
I had also hung the entire queue since early Friday behind it.
2. Some operations that might be done either on the printer or
in the computer are probably more efficiently performed on the
printer. For example, suppose you have a complex bitmap that you
wish to use repeatedly on a printer with local disk storage. It
seems that the all the PostScript operators for reading and
writing would come in quite handy. (Wandy has written to me
indicating that this may be included.)
3. Finally, it's not clear that the processor speed is always
the limiting constraint. I know of a good many Apple
LaserWriters and HP Laserjets (with PostScript cartridges) that
are still connected by slow serial links. I'm not convinced that
sending lots of simple PostScript over these links is faster than
sending a small amount more complex PostScript to be executed on
the printer itself.
Peter
|