- From: Robert Miner <rminer@geomtech.com>
- Date: Fri, 13 Nov 1998 10:25:43 -0600
- To: fiedorow@math.ohio-state.edu
- CC: www-math@w3.org, fiedorow@math.ohio-state.edu
> In another vein how efficient is MathML as a transport mechanism over > networks? I was wondering whether its verbosity might make it too slow for > those with narrow bandwidth to the Internet (sort of like Postscript). > How does it compare with HTML+equations rendered as images. To support Bill Hammond's assertion that even uncompressed MathML beats bitmaps, here is a data point. I just used WebEQ to translate $x = \frac{-b \pm \sqrt{b^2 - 4ac}}{2a}$ to MathML, and then create a JPEG, and then standard tools to hit a couple other formats. The results: LaTeX: 39 bytes MathML with all whitespace stripped: 219 bytes Indented MathML: 401 bytes quad.gif (no antialiasing for best compression) 1801 bytes quad.jpg (default quality) 2066 bytes quad.uu (uuencoding of quad.gif for mail, etc) 2617 bytes quad.jpg (highest quality) 3201 bytes > Also will might MathML's relative complexity lead to buggy implementations, > tending to a babel of mutually incomprehensible dialects? This is indeed a worry. The Math WG is trying to pulling together a test suite of MathML fragments (mostly contributed by the various implementation projects around) that all renderers should be able to handle -- not unlike TeX's trip test. Of course, that won't stop particular renderers from adding extra non-standard features, and other people writing to those features. On the other hand, for once, I believe MathML's verbosity and complexity work in our favor. Since MathML will primarily be generated by software tools, it is much, much easier to achieve a pretty good level of compliance than it is with a mutable, hand-authored language like TeX. In the reasonably near future, there should be a half dozen or so other major MathML packages to benchmark against. If your new MathML software is compatible with them, chances are it will be compatible with the rest. With software, the path of least resistance is towards regularity, and algorithmic uniformity, as oppose to hand authoring, where the path of least resistance is often toward quirky coding that is short and easy to type, or that appeals to the author for whatever whimsical reason. --Robert ---------------------------------------------------------------- Robert Miner http://www.webeq.com Geometry Technologies, Inc. email: rminer@geomtech.com phone: 651-223-2884 ----------------------------------------------------------------
Received on Friday, 13 November 1998 11:25:13 UTC