laser-lovers@uw-beaver (02/03/85)
From: Brian Reid <reid@Glacier> I've been tremendously busy this week trying to catch up with my coursework and keeping ahead of my students. I've watched quite a few messages go by that I'd like to comment on. AUTOMATIC SCALING OF FONTS I've read a few complaints that PostScript is bad because it scales fonts, and that good graphic design requires that fonts be scaled. I claim that this whole issue is a red herring. Here's why. If you are using an image description language that does not permit the scaling of fonts (such as ImPress), then you are required to specify every size of every font individually. If you are using PostScript, you are still permitted to specify every size of every font individually. PostScript does not take away any capabilities; it does not force you to use the same shape of letters for 4-point type as for 40-point type. It *permits* you to do that, but it does not *require* you to do that. There is a large class of typesetting activities, often referred to as utility-grade typesetting, in which scaled fonts from 8 points to 16 points (or bigger) will look just fine. For that kind of typesetting it is perfectly OK to use scaling. If you are working on the typesetting of some especially important document, and if you have access to fonts that are specially designed to be used in certain sizes, you can certainly use them. Since PostScript fonts are stored as outlines and not as rasters, you can in fact store a larger number of fonts in the same amount of memory before it overflows. I might add that almost all of the commercial typesetting machines with which I am familiar do some form of scaling of the fonts. The Compugraphic 8600 that we have at Stanford uses the same outline for all sizes from 4 point to 127.5 point. It does not even give you the capability to specify different fonts for different point sizes, because of the way its naming scheme works and because all fonts for it must be purchased from Compugraphic (and they do not provide alternate designs for different sizes). QUALITY OF AUTOMATIC SCAN CONVERSION Henry Spencer of Toronto complained that the PostScript-generated rasters were not as good as bit-tuned rasters. Lynn Ruggles of Stanford said that all serious type designers work with outlines, and use various automatic techniques for scaling and rasterizing them. Bruce Nemnich of Thinking Machines said he was quite impressed with the quality of the rasters generated by an Apple LaserWriter being demonstrated by Sun Microsystems. Clearly the issue of whether or not automatically-generated rasters are of sufficient quality is a subjective one, and therefore I don't think I should try to argue it publicly. Obviously I believe that the PostScript-generated rasters are top quality. The important issue for contemplation on Laser-Lovers is that people understand the difference between specification and implementation. PostScript is not a program, not an algorithm; it is a specification language. It is completely public-domain, its documentation is available to anyone who has $30 to pay for the copying costs. The company that developed PostScript, namely Adobe, has an implementation of PostScript that they are quite proud of, and in fact it is that implementation that is the company's product. That's why they are giving away the specification of PostScript. Any other company is perfectly welcome to do their own implementation of PostScript without infringing Adobe's domain. Competition helps keep the world interesting. In particular, it is completely reasonable to think of an implementation of PostScript that uses raster fonts rather than outline fonts. The Xerox Interpress format is somewhat similar to Adobe PostScript, but it uses raster fonts rather than outline fonts. It is also completely reasonable to think of an implementation of PostSript that uses stored bit-tuned rasters when they are available (standard sizes in standard rotations) and uses outlines to generate its own rasters otherwise. The important concept is that PostScript is a way of specifying what a page should look like, not a particular implementation of that specification, and that it does contain the expressive power to describe and use fonts in terms of outlines. I think that the history of computer science has shown us that it is a bad idea to adopt standards that are too tightly tied to the limits of current technology. PostScript is a page description standard that is not limited by current raster-based technology, and for which there is a pilot implementation (Adobe's) that demonstrates its feasibility even with today's technology. Furthermore, it is completely public domain and it is completely independent of the word size, two's complement properties, instruction set, or addressing of any particular computer, and efficient implementations have been demonstrated on several popular modern computers. My summary claim is that PostScript is the obvious choice for a standard for page specification, and that alternative implementations of PostScript are welcome to embody the theories of their implementors, such as the vital necessity of bit-tuned fonts or the importance of efficiency over generality. It is also perfectly reasonable to implement a PostScript subset, such as all of the imaging operators but none of the scaling, rotation, or halftone and grayscale stuff, and document or market it as a subset implementation. Certainly all of the different page description schemes that are in use today can be isomorphically transformed into PostScript subsets. Doing that would enable a common interchange standard for images, a shared set of image management software, and would reduce the need for special-purpose "drivers" at the back end of text formatting programs. -- Brian Reid decwrl!glacier!reid Stanford reid@SU-Glacier.ARPA
laser-lovers@uw-beaver (02/07/85)
From: Mendelson.es@XEROX.ARPA The purpose of this message is to correct an error about Interpress in the recent message from Brian Reid which states: "The Xerox Interpress format is somewhat similar to Adobe PostScript, but it uses raster fonts rather than outline fonts." That is a fallacious statement. Interpress makes absolutely no assumptions about how fonts are represented. It defines a very general mechanism for obtaining character shapes that includes both raster and outline definitions. An Interpress font contains, among other things, a set of composed operators (procedures) each of which when executed causes a character of a given shape to be imaged at a previously designated position on the page. Interpress makes no statement whatsoever about the nature of the composed operators contained in a font. They can contain outline definitions of characters, or bit-map definitions, or any other representation that anyone can create. What is true is that current commercially available Interpress implementations from Xerox use bit-map representations of fonts because of advantages in performance, speed and quality. It is my understanding that current Adobe implementations require one second of time to scan convert just a single outline-defined character. By comparison, the Xerox 9700 decompresses the bit-map fonts and prints a whole page in that one second. Brian Reid also makes some statements which impute capabilities solely to PostScript, but which apply equally to Interpress. I substituted the word "Interpress" for "PostScript" and "Xerox" for "Adobe" in the following statements quoted from Reid's message, and generated equally valid statements. I quote them here with our above defined substitutions indicated in square brackets: "The important issue for contemplation on Laser-Lovers is that people understand the difference between specification and implementation. [Interpress] is not a program, not an algorithm; it is a specification language. It is completely public-domain, its documentation is available to anyone who has [$xx] to pay for the copying costs." "The important concept is that [Interpress] is a way of specifying what a page should look like, not a particular implementation of that specification, and that it does contain the expressive power to describe and use fonts in terms of outlines." "I think that the history of computer science has shown us that it is a bad idea to adopt standards that are too tightly tied to the limits of current technology. [Interpress] is a page description standard that is not limited by current raster-based technology, and for which there is a pilot implementation ([Xerox's]) that demonstrates its feasibility even with today's technology. Furthermore, it is completely public domain and it is completely independent of the word size, two's complement properties, instruction set, or addressing of any particular computer, and efficient implementations have been demonstrated on several popular modern computers." "My summary claim is that [Interpress] is the obvious choice for a standard for page specification, and that alternative implementations of [Interpress] are welcome to embody the theories of their implementors, such as the vital necessity of bit-tuned fonts or the importance of efficiency over generality. It is also perfectly reasonable to implement a[n Interpress] subset, such as all of the imaging operators but none of the scaling, rotation, or halftone and grayscale stuff, and document or market it as a subset implementation. Certainly all of the different page description schemes that are in use today can be isomorphically transformed into [Interpress] subsets. Doing that would enable a common interchange standard for images, a shared set of image management software, and would reduce the need for special-purpose "drivers" at the back end of text formatting programs." -------- I suggest an objective evaluation of Interpress and Postscript and welcome the opportunity to participate in such an eveluation. I will continue to respond to questions and comments as suitable. Jerry Mendelson