laser-lovers@uw-beaver (01/26/85)
From: Mabry Tyson <Tyson@SRI-AI.ARPA> Thanks for the good description of the differences between PostScript and Impress. I have a question though. A significant portion of our printing is bitmap dumps of 1Kx1K screens. Since you say that the Impress description is characters, how much bigger would the description of the screen be (assuming it is fairly dense in turned on bits). The other question would be in how flexible is the description language. If I have a complicated display (not generated by some simple plotting commands), would the only reasonable way to send it to a PostScript printer be the equivalent of sending a bitmap? -------
laser-lovers@uw-beaver (01/26/85)
From: Mike Caplinger <mike@rice.ARPA> Perhaps I am confused, but Brian seems to be saying that PostScipt uses a size-independent outline description for a character in a particular font, and can arbitrarily rotate it before conversion to a bitmap. I was under the impression that there was no such size-independent representation. Isn't that what the whole hoopla over bit-tuning is all about? Certainly MetaFont is incapable of producing a uniformly great font at 300 dpi resolution, 10 point. We have used both a Xerox 2700 and an Imagen 8/300, both 300 dpi, but the character of one-pixel line drawing on the two engines is so different that CMR fonts that look great on the Imagen were awful on the 2700. Or is the outline description some incredibly sophisticated representation that takes all into account? - Mike
laser-lovers@uw-beaver (01/26/85)
From: Robert Morris <ram%vax2.uucp%umass-boston.csnet@csnet-relay.arpa> I read Brian's thoughtful and clear comparison after sending my mail about font quality and matching problems. His explanations call for no enhancement, but his conscience attack is contagious. I neglected to mention that I have a commercial association with Interleaf, Inc. which is both a buyer and seller of printing systems, including from Imagen. We are regularly evaluating printing systems from many vendors. I believe companies like Interleaf will be induced by the marketplace to support some or all standard page description languages. Personally I hope the forces tend toward PDL's as competent as PostScript and Impress. Among those I see people potentially demanding as well, Interpress comes to mind if only it had more graphics in it. Many people seem to want variants of some of the standard graphics metafiles like VDM. The next year or two will prove tumultuous and interesting I think. bob morris umass-boston dept of math and cs and Interleaf,Inc.
laser-lovers@uw-beaver (01/26/85)
From: Brian Reid <reid@Glacier>
Perhaps I am confused, but Brian seems to be saying that
PostScipt uses a size-independent outline description for a
character in a particular font, and can arbitrarily rotate it
before conversion to a bitmap.
Nope, that is what I am saying. You are not confused.
I was under the impression that there was no such
size-independent representation. Isn't that what the whole
hoopla over bit-tuning is all about? Certainly MetaFont is
incapable of producing a uniformly great font at 300 dpi
resolution, 10 point.
It may or may not be impossible, but the PostScript printers do it and
do it well. If you want to see an example, look at the recent POPL
conference publicity. The flyer announcing the conference announcement
was set with Scribe on an Apple LaserWriter in 8-point Times Roman
using a font described with ordinary outlines. One of the recent issues
of CACM (I think it was the November issue) had a 2-page advertisement
for the POPL conference somewhere towards the back. That advertisement
was also set on an Apple LaserWriter in PostScript. It isn't quite so
remarkable because it uses 10-point letters and does not use any
rotated fonts or graphics, but it is an example of the PostScript
raster conversion. Remember that it has been through a printing press.
I have carefully avoided learning anything about how the internals of
the PostScript font mechanism work, because I often have difficulty
keeping my mouth shut and I don't want to give away any Adobe secrets.
But you may rest assured that an ENORMOUS amount of software cleverness
and expertise in the nature of laser printers has gone into the
PostScript system. I think that I will have to doublecheck back with
the folks at Adobe to find out what I am allowed to say and what I am
not allowed to say before I give you any more information about the
font representation, but please understand, for the moment, that the
fonts are indeed represented as size-independent outlines and that they
can be artitrarily scaled and rotated before being scan-converted and
that the scan-conversion works just fine in point sizes down to 4 on a
300dpi printer.
Brianlaser-lovers@uw-beaver (01/26/85)
From: Clive Dawson <CC.Clive@UTEXAS-20.ARPA> Brian-- One of the things you mentioned in your message was that Impress has the ability to deallocate memory for no-longer-needed fonts. This works quite well on the old Imprint-10 printer, but has apparently been disabled on the new 8/300's. The result is that glyphs will continue to be down-line loaded to the printer until it runs out of memory. After that, you lose; any new glyphs that must be sent will simply show up as small boxed question-marks (i.e. undefined glyphs). The story I got from Imagen on this is that the 8/300 is actually running two processes-- one which receives the data from the host and stores the glyphs and does some of the page setup, and another which sends the pages off to the print engine. If you watch the 8/300's terminal output, you will note that that the first process can actually be several pages ahead of the second process. In any case, deleting glyphs turned out to be too hard because there was apparently no way to know whether the printing process was really finished with a glyph or not. Our users have quickly learned that long documents which use a variety of different characters in non-standard fonts, even if few and far between, have a high probability of eventually losing. The solution is to split up the document and print the pieces in several different jobs. One reason we may notice this more than others is that we ordered our 8/300 with a serial interface which comes standard with 512K bytes of memory. If you order an Ethernet interface, enough extra memory comes with the printer that running out of glyph space is apparently much less common. We will probably be ordering more memory for our printer anyway, which will not only help this problem but also allow us to print full bit-mapped screen images from our Symbolics 3600 Lisp Machines. At the moment the printer won't even hold half the image before running out of room. Cheers, Clive -------
laser-lovers@uw-beaver (01/30/85)
From: ihnp4!utzoo!henry@uw-beaver.arpa > It may or may not be impossible, but the PostScript printers do it and > do it well... > > ...fonts are indeed represented as size-independent outlines and... they > can be artitrarily scaled and rotated before being scan-converted and > ... the scan-conversion works just fine in point sizes down to 4 on a > 300dpi printer. "Beauty is in the eye of the beholder." The more experienced typesetting people at the Typesetting BOF at Usenix Dallas said that the scan conversion does *not* work "just fine", and that the loss in quality (compared to fonts with separate masters for each size and orientation) *is* visible in output from the Apple printer. Quite probably it's still impressive (I haven't seen it yet), but Adobe not has found a magic way around the problems of size-independent description. Henry Spencer @ U of Toronto Zoology {allegra,ihnp4,linus,decvax}!utzoo!henry
laser-lovers@uw-beaver (01/31/85)
From: Bruce Nemnich <godot!bruce@cca-unix> At the Usenix typesetting BOF, I was one who was pretty skeptical about scaling fonts from one master description over such a wide size range at this resolution, but I hadn't seen the output yet. I spent some time at the Sun booth the next morning, and I came away very impressed. They evidently have some very good algorithms for generating well-tuned fonts at medium/low resolution; I found it hard to believe they were generated without hand tuning (especially at < 6pt!). I have always been a big believer in disproportionate scaling of stroke thickness and glyph width, especially at medium-to-low resolution since some engines do a lousy job with fine strokes and it is *so* difficult to generate optically correct lowres rasters, but the quality of Adobe's fonts does not degrade nearly as much as I expected at small point sizes. I have done some low-resolution work with the old metafont, and it is difficult to get reasonable results without hand tuning. I am happy to hear they are sensitive to the differences among engines and tune their algorithms to each. Halftones are another difficult problem with which they seem to have done quite well, though I didn't see too many examples. I am probably going to buy one of these things instead of the QMS 800 I planned to get; it's so much more "the right thing" (cheaper, too!). The word I got from QMS is that they have about a half-dozen beta sites for the 1200A. The 2400 will be the next to be converted; they haven't committed to doing a version for the 800. --Bruce Nemnich, Thinking Machines Corporation, Cambridge, MA ihnp4!godot!bruce, bjn@mit-mc.arpa ... soon to be bruce@tm.arpa