Newsgroups: comp.databases,comp.lsi,comp.parallel.pvm,comp.parallel.mpi,comp.org.acm,comp.org.ieee,comp.protocols.misc,comp.realtime,comp.arch,comp.software-eng,comp.sys.super,comp.theory,comp.dsp,sci.math
From: Mike Curtis <curmi02@unix9.ingres.com>
Reply-To: curmi02@unix9.ingres.com
Subject: Re: Publishing Scholarly Work on the Web -- opinions?
Organization: Computer Associates International
Date: Fri, 11 Oct 1996 11:24:49 +0100
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-ID: <325E2071.7990@unix9.ingres.com>

Bernd Paysan wrote:
> 
> J.Sigbrandt wrote:
> > I think everyone using computers should be aware of these issues.  I for one
> > now insist that the programs I use store my work in the human readable ascii
> > format.  It would be difficult to convince me to use a program with a
> > proprietary electronic data format.
> 
> The only option from plain ASCII IMHO is LaTeX then. Unless you do
> complicated things, it's still plain ASCII and human readable. Yet it
> covers all the information needed for text processing, and it is the
> only text format I know that is close to be stable for more than 10
> years, which isn't true for HTML (not even close).
> 
> --
> Bernd Paysan
> "Late answers are wrong answers!"
> http://www.informatik.tu-muenchen.de/~paysan/

LaTex has been around a lot longer than HTML so the comparison is not
fair, and anyway it does not answer the original question (well not
quite the original question, that was so long ago that we can no longer
retrieve it!) because having LaTex on a floopy disk that can no longer
be read does not help.

There cannot be a medium that we can guarantee will last forever and for
which the technology to read from it will also last so the only possible
answer to the problem of indefinite storage is to transfer information
from one format to another and from one format to another when
necessary. We can make this process easier by considering the documents
that we want to save as objects rather than just files of text. The
methods for these objects include reading and replication. The
replication method makes it relatively easy to change media, before a
medium becomes outdated the method is called for each object being
stored naming the new medium. The next problem is the format of the
methods, storing 80*86 executables is limited by the (hopefully short)
lifespan of processors running this instruction set. Interpreted or
semi-interpreted languages would be better so we can hope that an
interpreter could be available for a significant length of time
particularly if source code is included and clear instructions on how an
interpreter can be built. In fact it should be possible to use the
bootstrap principle to start with a very small program that it should be
possible for anyone to build from some very simple instructions that can
be expressed mathematically so we do not have to rely on the ability to
read English.

I seem to be getting back to the Universal Turing Machine here!

We can then go a stage farther and realise the type of computer that
will be able to run our interpreter might not exist so we have to
include instructions on how to build one. Takes a lot of space but this
all has to be done only once per library.

regards
Mike


-- 
Mike Curtis               tel: +44 1753 679935
curmi02@unix9.ingres.com  fax: +44 1753 679900
Computer Associates International

