On Thu, 2005-05-19 at 18:47 +0100, Antonio Carlini wrote:
Jim Leonard wrote:
Which means they can run on any machine with a C
compiler. So
what's with all the paranoia? Just use whatever works as long
as more than one major platform can extract it.
What's C? Will that be less dead in 100 years than Algol is today?
I know Algol's not completely dead - but it's certainly declining ...
How long before C is no longer available everywhere? How long before
your source no longer compiles without significant effort?
I have no problems with including a ready made unarchiver, but you
also must include a textual specification of the archive format.
Why not include full specs for the computer the data is supposed to be
on as well as how to build the computer -- and don't forget to include
an ascii chart and an english dictionary so they can figure out the
instructions!
The format issue suffers from the same problem as the media issue does:
there is no such thing is an eternal format. To put it simply: its the
content, stupid.
We're not building a time capsule (to everyone who's been complaining
about data accessibility 500 years out). I see it as an archive in the
LIBRARY sense of the word. Who is the target audience, really? Is it
some future generation (only) which might have no interest in CP/M -- or
is it collectors NOW who would like to preserve the content for
themselves and the future? I think its the latter.
The bundling file formats are not going to ever be designed for the
future correctly. Ever. We should have something that is useful today
and upgradable in the future. Once the data is extracted and collected
into an archive, "updating" it to whatever format the future uses will
be a simple as:
for n in *.zip; do
mkdir temp
cd temp
unzip ../$n
newarchive ../whatever.new .
cd ..
rm -rf temp
done
(hehe, since unix and sh will live forever :)
The same holds true for the meta data portions of the package. If we
use XML or plain text or whatever, its a simple matter of programming to
convert it to the Format of the Future(tm).
The trick is to make it comprehensive enough to provide the information
which may be needed later and easily updated for future systems.
Brian