[pdftex] TeX as a composition server?

[Jeffrey McArthur]

The composition systems I created typeset directly from XML. There is no
need to convert XML to TeX markup.

Again, if you are willing to work with active characters in TeX, there is no
reason to worry about font limitations. Using UTF-8 encoding, and setting
some characters active to handle the UTF-8 escape sequences I typeset the
Library of Congress Subject Matter headings. I had to set up several virtual
fonts (I like VF files). The Library of Congress Subject Matter was
particularly difficult because Unicode does not include all the glyphs
needed (the transliterated Vietnamese are not included in Unicode 4 or 5).

Jeffrey McArthur
cell: 610-389-0734
home: 610-450-6115
email: jeffmcarthur at comcast.net
http://www.jeffreymcarthur.com
http://twitter.com/JeffreyMcArthur
http://jeffmcarthur.blogspot.com


On Tue, 26 Oct 2010 23:38:28 +0200, you wrote:

>On Tue, Oct 26, 2010 at 11:20 PM, Reinhard Kotucha
><reinhard.kotucha at web.de>wrote:
>
>>
>> Before you decide to use plain TeX, please investigate whether it
>> provides everything you need.
>
>
>Well, the very first thing to determine is whether TeX can possibly handle
>the throughput I require.  I realize this is a very unrealistic test for a
>number of reasons, but I just want to see the upper bound.  On my laptop,
>running a Lorem-ipsum loop, I seem to get about 8000 pages per minute.  This
>is a dual-core Intel 2.2GHz machine, so I'm assuming I could at least double
>that on a newer, more muscular system.  That just about the minimum to keep
>some variation of TeX as a candidate for this job.  I'll have to try this
>test on a server-class machine to get a better indication.
>
>Now to get more realistic, I have to account for generating TeX mark-up from
>XML, handling placed figures, (no floating), multiple fonts, languages and
>eventually writing directions, I'll have to build more realistic prototypes.
> Also, this will actually be more of a succession of short (8-12 page)
>documents, rather than one long one (though it could be processed as one
>long one).  I've read about some work with making a TeX daemon that would
>amortize the startup costs over the entire life of the process.  But there
>would be no point in doing these experiments if I couldn't even get plain
>TeX up to speed.
>
>Thank you for you thoughtful responses.  I realize there are lots of options
>and trade-offs in picking a TeX variant and format.  That's how I expect to
>be spending my next week or so.
>
>Thanks!
>
>-pd