Re: Problem with dvips and Acrobat

["Berthold K.P. Horn" <bkph@ai.mit.edu>]

Hi:

At 02:02 AM 98/12/05 +0100, Yannis Haralambous wrote:

>I had a very curious problem. Some of you may know my yhmath
>package and the font with the very large math accents. I recently
>did a PostScript version of it. It printed and previewed (gs)
>without the slightest problem. When I tried to distill such a
>PostScript file, Acrobat refused it. After a long search it came
>out that the problem was in the metrics of the font:

>As you can see, the widths of Yhcmex characters start at
>1209. Well, all characters with width greater or equal to 720
>make Acrobat Distiller abort!

There are some limits on numbers in Type 1 fonts.  The black and white
bible says on page 26 that the limits are -2000 to +2000. Different
interpreters however, exceed the limits by various amounts.
All interpreters seem to be able to handle at least 4096 (on the
1/1000 em scale).

There are also some limits on numbers in PS interpreters. And again,
some are better than others.  Some QMS interpreters e.g. have
lower limits than Adobe interpreters.  Which leads to problems when
you work with say scaled points of which there are 65536 per pt.

By the way, when you use the Distiller you really see the odd nature
of DVIPS's redefining the font metrics: it simply makes no sense.
Obviously DVIPS does not know the resolution of the final output,
so any grid fitting it does is not only not a help, but actually makes
things worse (that is, as often as not it is adjusting the actual width
in the wrong direction).  And basing the grid fitting on the grid size determined
in PS would do no good, since the PS code would see the default 144 dpi
grid of Distiller, not that of the final output device.  This aside from the
fact that PS code should be resolution independent, not tuned to a
specific resolution...

>I suppose that these are PostScript units, and hence 720 means
>one inch. Can it be possible that Distiller cannot treat
>characters wider than one inch, in 10 points fontsize?

Part of the solution most likely is to scale the font so it does not 
exceed the limits (then compensate by changing the FontMatrix of course).
We had to do this with some of the math extension fonts.
Sun NeWS (thankfully now dead) had a big problem with these.

Also, you may want to use a coarser coordinate system in the PS
code to keep the numbers smaller.  We had to do this for some
Harleqiun based PS RIPs which cannot handle large numbers in PS.
Of course, they said it was silly to want a million point font, but this
is nonsense, since a million of whatever units you happen to be
using can be a small number of points.

You can easily test these two different concerns by changing the
point size at which the font is used and chanigng the scale of the
coordinate system you are working on.

Regards, Berthold Horn.



Berthold K.P. Horn
Cambridge, MA		mailto:bkph@ai.mit.edu