NetBSD-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: kerTeX (unices/Plan9): dvips(1) added.


On Mon, Dec 06, 2010 at 07:19:06PM -0500, Christopher Berardi wrote:
> I'm interested in your project. But, I have a few questions. 1) What is the 
> general rationale for your project (what goals do you hope to achieve and 
> what prompted you to begin). 

To get to the truth of TeX, since the truth about a subject is the 
shortest path to it.

It took me less time to start from scratch than to try to get more than
one gigabyte of "things" to find what I need or to try to "port" the 
current TeX distributions.

D.E. Knuth has made TeX/METAFONT so that his written work will not
anymore depend on external things---printing technology ; fontes
---. He wanted to not be "booknapped". And the result is that his
software has been "softnapped".

To have TeX now, you have to search a needle in a haystack. I wanted the
needle, so I have burned the haystack.

Furthermore, TeX has been erased for the public since it is just a small
amount in terms of bytes of the whole. Typical LaTeX user says : "We
don't need TeX! We use only LaTeX!" not knowing that LaTeX are just
macros compiled/interpreted by TeX ! LaTeX users have been made TeX
illiterates---they are on the very same ligue as a PC user asked "What
is your operating system?" and answering : "Office"---.

2) How does it perform compared with, say, texlive's plain TeX in areas such as 
cpu consumption, memory usage, execution time, etc. (especially when document 
size gets large, like to 5-10k pages) -- do you have any benchmarks? 

To make benchmarks you need to install first... Texlive. Since I have
no systems able to run it ; no gigabytes of space to waste ; no days of
download to waste neither, the benchmark will be a theoretical one.

kerTeX uses the latest versions of TeX and al. This is, for the core
engine, the very same code used by all other TeX distributions (except
that the installation of the current TeX distributions is such a
nightmare that the current installations are probably not as up to
date as kerTeX), since it is Knuth's (and some others for auxiliaries)

The translation from WEB/Pascal to C is done with the same original work
from Tomas Rokicki---I have made more work on this than has been done by
the "community" in 20 years; mainly some cleaning and simplifications---.

Hence, for the core it is the same. But since kerTeX depends only on
C89---this answers your third question---it should run faster than
all other versions.

As for the size of the documents (more importantly the size of the macro
packages since once a page is ejected it is forgotten), it is tuned
via change files, hence kerTeX is also producing "big" versions
able to chew big macro packages (typically LaTeX).

The only area where some improvement in speed is possible---tuning
reintroduced in kerTeX with the addition of John Hobby's MetaPost---is
the possibility on some hardware architectures (or C implementations)
to speed some arithmetic routines by using floating point versions.
But this concerns only METAFONT and MetaPost, not TeX; and will be
not set by default since the whole TeX system (TeX and METAFONT)
is a compiler and should be integer based by default.

So to summarize: kerTeX should run on more systems (in fact on all
systems having a C compiling framework, including via cross compilation)
and use less ressources, in space and time, than other versions;
and should be able to support every correct TeX macro packages---for
example LaTeX---and should produce the correct result since kerTeX
passes the TRIP and TRAP tests.

        Thierry Laronde <tlaronde +AT+ polynum +dot+ com>
Key fingerprint = 0FF7 E906 FBAF FE95 FD89  250D 52B1 AE95 6006 F40C

Home | Main Index | Thread Index | Old Index