Vadim V. Zhytnikov wrote:
>Konovalov, Vadim Vladimirovich (Vadim)** CTR ** writes:
>
>
>>I suggest to not include generated files into the distribution.
>>So it would be very great if PostScript and PDF files are removed, just because they're generated out of TEX.
>>All of these could be available from elsewhere.
>>
>>But even of more concern is duplicating a lot of files in, for example, "es" and "es.utf8" directories.
>>Freshly unpacked 5-11-0 distribution contains 94 binary identical files within these two directories, of more than 3.5 MBytes of wasted space.
>>Same thing for "pt" subdirectories.
>>With all the respect to translated material, what the need for duplication?
>>
>>Is it possible to generate non-UTF8 out from UTF8, or vice-versa?
>>
>>I can write some 10-lines Perl script to do the encoding. Is it possible to use Perl within makefiles?
>>
>>
>
>First of all I don't understand why to care too
>much about size of _source_ distribution.
>For 5.11 it is 14Mb of which 1Mb is es.utf8/pt.utf8 overhead.
>Is this really too much? Yes it is more than this in
>unpacked sources but who keep unpacked sources for
>long time.
>
>
IMO the source distribution should contain as little derived files as
possible, for the following reasons.
- distribution without generated binaries will provide better impression
about its quality. Its like having files like README.txt~ in some distros :)
- size isn't the very improtant argument nowadays, but still adds to the
overall picture
I just looked into CVS at
http://maxima.cvs.sourceforge.net/maxima/maxima/doc/info/ and see that
these files are not duplicated.
(execpt *.gif, *.ps.gz and pdf files inside /maxima/doc/info/figures/)
Why shouldn't source distribution be like CVS version?
Its okay for binary distribution to be stuffed with all PDFs and all,
but it seems unnatural to me to have source distribution be in any way
larger than CVS.
In other words, I very much like to have CVS snapshot at the moment when
source files distribution was created. (with bigger one to be like
"externded" source distribution)
Recently I was impressed on how full MAXIMA coudl be run within 2 files:
lisp.exe/lisp.run and maxima.mem
Very impressive and brilliant on having powerful 2-files CAS!
>On the other hand lang + lang.utf8 is indeed an overhead
>and it would be nice to eliminate it.
>Your guess is right - the motivation is to make
>installation form source tarball Perl-free.
>We don't need Perl to recode files we need Perl
>to create help system index files for .info and .html
>documentation.
>
>
>
Attached is the CVS patch to INSTALL.win32 file. I intended to fix the
typo, opened the file and touched several places to slightly improve
explanations.
However more editing of this file needed:
- currently the file explains only GCL way of build, one could also want
to build with CLISP (this is appears to be easier), but currently the
file doesn not even mention different LISP implementations.
- Some phrases are messy. My English is not good, but may be I'll
eventually add some improvements.
Vadim.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: INSTALL.win32.diff.txt
Url: http://www.math.utexas.edu/pipermail/maxima/attachments/20070331/ab0a6964/attachment.txt