- From: Raph Levien <raph@acm.org>
- Date: Fri, 3 Dec 1999 15:08:17 -0800
- To: www-svg@w3.org
Hi svg'ers, I've been continuing to research font formats since my last screed, largely in the context of design work for a new X extension for advanced 2D graphics rendering. I thought I'd bring a heads-up to you guys. First, there is a lot of variation in rendering for hinted type1 fonts. The Adobe renderer is a lot better than either of the two major free implemntations, GhostScript and the one in the X server (which is derived from donated IBM code). I was under the impression that the free ones were better, so I was disappointed. Just a word of warning. A comparison is available at: http://www.dcs.ed.ac.uk/home/jec/programs/xfsft/renderers.html Second, I came across a paper that explains how to do _awesome_ automatic hinted rendering. This is paper 89/2-17.ps in John Hobby's bibliography (a rather impressive collection of documents, by the way): http://cm.bell-labs.com/who/hobby/pubs.html This technology has some very interesting properties. First, it doesn't depend on manual hints in the font; it works totally automatically. This is significant because the current SVG font proposal has no hints. Second, while the paper doesn't explicitly discuss the issue, it would appear likely that it can be adapted well to rotations and affine transformations. The constraint subsystem already contains an "integer offset property" mechanism for getting smooth, consistent results from diagonal lines. Third, it seems to be very nicely script-independent. Type1's hinter contains a number of features optimized for latinlike and kanji scripts. There is some interest within the free software community (Gnome and X) in doing a production-quality implementation of these ideas. It looks like they'll require quite a bit of investment of time and mathematical skill. If there's anyone out there who'd be interested in funding such an implementation, I'd be happy to hook you up. It would appear that the techniques in this paper are free of patent problems. Certainly John Hobby is unaware of any such issues. Third, I've come across some patents relating to antialiased + hinted fonts, ie using the hints to perturb the outlines, then filling them using an antialiasing technique. Microsoft has a patent (US 5,684,510) that would appear to cover this, as does Apple (patent number not handy right now, but I can dig it up). Thus, it seems like requiring antialiasing with hinting is probably not a good idea. Since some people might like to use it if it were available, it might be putting in an option, much like the existing one for selecting antialiasing. Hope some of this information comes in handy. Raph
Received on Friday, 3 December 1999 18:10:10 UTC