Notes on producing the CD-ROM proceedings
.. as learned during the IEEE 1997 Mohonk Audio Workshop (WASPAA'97)
by Dan Ellis <dpwe@icsi.berkeley.edu>
1997nov20
- Jont disagrees, but I think PDF is the way to go. Converting the papers
to PDF was a good way to shake out any problems in weird postscript, and
the size savings were sometimes dramatic. Jont doesn't like the Acrobat
user interface, but I think it's fine.
- One of the big advantages of PDF is that pretty nice full-text indexing
is provided for free within Acrobat. It does require a little bit of work
to set up the index, and you have to be using the version of Acrobat Reader
that includes the search resources (it's free, though).
- I bought Adobe's Distiller for Unix to do PDF conversion. This works
pretty well, but it quietly omits some of the useful functions available
on the Mac or PC versions. In particular, Acrobat Catalog, the separate
program that generates the indexes mentioned above, isn't included on Unix
(fortunately, we also had bought the PC version of Acrobat, so I could
build the index there). Acrobat Exchange (the editing version of Acrobat
Reader) has the ability to make small changes to the text on a page on
the Mac and PC, but the Unix version inexplicably lacks this feature.
- Displaying papers at screen resolution looks a million times better
if the fonts are 'hinted' (i.e. pixels on the edge may be shaded gray rather
than just black or white). Acrobat will automatically hint documents if
they use fonts that are encoded as "type 1" (scalable) (common
fonts like Times and Helvetica are usually not encoded in the file at all,
so Acrobat uses its own, hintable, versions for these fonts). However,
of the papers we received, a minority included their own "bitmapped"
(non-scalable) fonts, which were illegible under Acrobat until you zoomed
in pretty close. Some of these were weird Windows drivers (which I can't
comment on), but mostly they were LaTeX documents which used the Computer
Modern fonts. It's quite easy for people to generate these documents with
scalable fonts, but they have to (a) download and install the scalable
(outline) versions of the CM fonts in the widely-available free "BaKoMa"
package, then (b) use the appropriate flags to dvips, which I think is
-O. There's a reasonable FAQ on this topic on the Adobe web site, and elsewhere
(referenced in pdf-prep-notes).
- I should probably have gone back to the authors of the papers with
bitmap fonts and had them regenerate their PS, but I wasn't publications
chair and I didn't want to get that involved. Instead, I used a free Unix
program, which I learned about from the Adobe web page, called "substitute"
that guesses from the bitmaps which fonts are actually being used, then
rewrites the dvips-generated postscript to use scalable fonts instead.
This is a highly heuristic process, and sometimes it substitutes the wrong
font which looks really horrible, but on the whole I found it to be very
useful. Still, avoid using it if you can (more details in pdf-prep-notes).
- In the end, we printed the proceedings from a combination of all the
PDF files, which I made by running all the PS files into distiller in series.
I had to disable the paper-size selecting commands in the postscript interpreter
to make this work (postscript included in pdf-prep-notes).
The problem was that a large number of papers were formatted for A4 paper,
which meant they had silly margins when forced onto 8.5x11 paper. However,
by rendering the master PDF on larger paper, then going through
selectively changing the margins - which you can do per-page-range in Acrobat
Exchange - I was able to get a single PDF with reasonable centering for
all files.
- You'll notice in the 1997 proceedings that many of the images - particularly
screen shots - look washed-out and fuzzy. This is because Acrobat treated
them as "RGB" images i.e. photographs, and resampled them to
72 dpi then jpeg encoded them. This was an unfortunate mistake: By using
"-colorres 300 -grayres 300 -colorcomp rle -graycomp rle" to
select lossless run-length encoding of color and grayscale images resampled
to 300 dpi, we could have retained full quality for these rasterized images.
So, don't make the mistakes I did...
- I prepared the PDF index in FrameMaker5 for the Mac. The big advantage
of using Frame is that it has support for hyperlinking built-in, which
it somehow stores in the postscript it writes (this is a check-box on the
print dialog), so that when you later run Distiller to generate the PDF,
it already has all the hyperlinks. This was particularly useful for the
author index, which was automatically generated, once I had gone through
marking each author's name in the text with an "index" entry.
Back to Mohonk WASPAA'97 hints & tips