Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!rutgers!rochester!pt.cs.cmu.edu!cadre!pitt!cisunx!jbw
From: jbw@cisunx.UUCP (Jingbai  Wang)
Newsgroups: comp.text
Subject: Re: New dvi2ps with on-line help and continuous font sizes
Message-ID: <17960@cisunx.UUCP>
Date: 7 May 89 18:14:38 GMT
References: <17892@cisunx.UUCP> <1989May3.204353.29194@cs.rochester.edu> <17924@cisunx.UUCP> <1989May4.154256.22958@cs.rochester.edu> <17943@cisunx.UUCP> <1989May5.230959.19926@cs.rochester.edu>
Reply-To: jbw@unix.cis.pittsburgh.edu (Jingbai  Wang)
Distribution: usa
Organization: Univ. of Pittsburgh, Comp & Info Sys
Lines: 86

In article <1989May5.230959.19926@cs.rochester.edu> ken@cs.rochester.edu (Ken Yap) writes:

>The tfm files are only about 1k each and there are only 10 or 20 of them.

Once tfm are preloaded in *.fmt, there is no importance for its existence
unless some smart users would use them by \font command. The key point is
that for a single TFM file, there need to be a dozen of PK files at different
mag's for device drivers to use.

>A whole crew of SAs? Oh dear, I must be underpaid. :-) To make a slitex
>after having made the initex and virtex binaries,
...
>All of 3 minutes.

Supporting does not solely mean setting up the binaries and path. In
a university as big as ours, there can be thousands of users, smart, dull
or in between. They ask you all kinds of questions, especially because
they can get nothing out of Knuth's books. Consulting is much more
important here than maintenance. It is why I was asked to write TeXGuide.tar.Z.

>No disagreement here. Turning to more interesting topics perhaps you
>can tell us how you smooth the scaled bitmap fonts (I assume you do).
>It would be a nice thing for future dvi2ps's to have.

Actually I have only done the best to the availability of time, for the
time being this version of dvi2ps does not smooth the scaled bitmap fonts,
instead, it finds the bitmap font of closest size to reduce scaling factor
to minimum possible. As far as I can see, the results are so far so good.
There was a conference here, people try to make Posters 

MEN's 
-->

Ladies
<--

the font size should be no less than 480pt. Then this version of dvi2ps
was not ready yet. I told them to forget about TeX, use Scribe instead. In
this particular case, smoothness is not so important.

>Sideline: There was a lot discussion on TeXHax a while back on METAFONT
>vs PS. What more or less emerged from the dust is that one can see the
>difference in quality between MF's tailored fonts and PS's optically
>scaled fonts if one is fussy. Indeed, I can see some ugliness in the
>lower point sized PS fonts. But the pressure of wanting to save on disk
>space is in favour of PS fonts at the moment.  It remains to be seen if
>some new storage proposals can retain the quality of MF fonts with a
>single font description, a la PS.


Sure, I remember that, and I actually critisized it. Adobe could have
suited them for spreading their unthoughtful remarks. Actually most of
Adobe fonts are not bitmap fonts, they are either outline fonts or
stroke fonts (like Courier). Scaling is done analytically (by computing)
the contours first, and then fill the interiors with grids (instead of
dots). If you want to see more primary fonts of this type, check
out DISSPLA and TEL-A-GRAF vector fonts (when you use a plotter with
pen movements, vector fonts are the best) which in turn appeared in Turbo
Pascal (exactly the same vectors as far as I can see). Of course, the
plotter do not fill the interiors, but printers can (and so can the
screen display as in Turbo Pascal). Adobe fonts are not optically scaled as
would a Xerox copier. MF uses about the same principle. In other words,
cmr10.300pk, cmr10.360pk, cmr10.432pk, ..., are generated from the
same description of font outlines, but the dimensions are different and
the bitmap information is different.  Some people said Adobe fonts
are linearly scaled (and hence ugly), and MF is not (and thus beautiful).
I could not see any non-linearity from cmr10.300pk to cmr10.360pk, but
of course, cmr10.300pk and cmr12.300pk are not proportional to each other.
Well, I do not feel it is the right comparison, this is really like 
comparing Times-Roman with Helvetica, because strictly speaking cmr10
and cmr12 do not have the same font style.

Returning to the smoothing issue, there are standard methods
to smooth curves (see Numerical Recipe) that can be used to smooth
font contours, but it is not smart to pick up the data from PK files
to make such outline fonts. Instead, one has to come one step back to 
MF itself.

I personally use Scribe to write everything, because font support is
good and I have my own version of database. I once communicated with
Leslie Lamport for a while, he does not seem to favor TeX fonts either and
he advised me to devote time to support Adobe fonts instead. When you
feel that the small Adobe fonts appeared ugly, that's because you are
misled by those messages.

JB