>>>> "Jerry" == Jerry James
<james(a)eecs.ukans.edu> writes:
Jerry> On 19 Feb 2001 at 20:29:59 +0000, Jan Vroonhof
Jerry> <jan.vroonhof(a)ntlworld.com> wrote:
> It does as an extra mechanism. However I don't think you can
get rid
> of the index. You would lose the descriptions and the possibility to
> add run-time 'requires at some point in time.
Jerry> Hmmmm.... you'd also lose the md5 checksum. On the other
Jerry> hand, maybe there's a seed of a good idea in here
Jerry> somewhere. The current package-index file contains
Jerry> information with varying volatility. I'd group it
Jerry> something like this:
Jerry> Very stable (changes less often than, say, once a year on
Jerry> average): standards-version, maintainer, distribution,
Jerry> priority, category, dump, description, provides, type
Jerry> Somewhat volatile (between the other two): requires
Jerry> Highly volatile (changes with every package release):
Jerry> version, author-version, date, build-date, filename,
Jerry> md5sum, size
Jerry> How about an approach like this?
Jerry> - Maintain a package-info file that contains everything that is not
Jerry> highly volatile in the current package-index file. On doing a
Jerry> list&install, first check whether this file has changed, and download
Jerry> a new copy if necessary.
Another gripe about package-index is this:
Newly built packages get added at the beginning of the package-index
file.
When the package-index is loaded, the last (later in the file)
definition wins. This is broken as designed.
Jerry> - Name the packages themselves like this:
Jerry> name-version-authorversion-md5sum-pkg.tar.gz. (That will make for
Jerry> very long filenames, but humans shouldn't be directly accessing these
Jerry> files anyway.) Then use a parser like the one in my previous message
Jerry> to pull out version, author-version, filename, md5sum, and size.
Jerry> Putting date and build-date in there too is probably overkill, but
Jerry> I'll let somebody else decide that.
Jerry> What do you think?
Jerry> --
Jerry> Jerry James
--
Adrian Aichner
mailto:adrianï¼ xemacs.org
http://www.xemacs.org/