[Mingw-users] Re: OT: packaging alternatives

[Soren A <sor...@fa...>]

Cristian Adam <dr...@gm...> wrote in
news:169...@gm...: 

> Hello All,
> 
> I wish that for the big packages (gcc, bin-utils) there sould be
> some archives different from tarballs, some kind of smaller archives,
> I've done some tests with 7zip (www.7-zip.org, free, sources with
> LGPL license, the best) and I come with the following results:
> 
> gcc-3.2-core-20020817-1        7204k -> 3564k (49%)
> binutils-2.13-20020808-1       5363k -> 1303k (24%)
> binutils-2.13-20020808-1-src  12078k -> 7276k (60%)
> w32api-1.5                     1165k ->  606k (52%)

`tar' doesn't do any compression itself. `zip' and things like it
combine the function of compression and the function of archiving in one
application, whereas tar is *just* an archiver. The most recent standard
settled on for many Open Source releases is bzip2, which gives
significantly improved compression ratios over the traditional gzip
tool. 

What you are requesting is to add another very new combination
compression/archiving tool format to the formats (tar.gz, tar.bz2) that
newbies and folks who haven't worked with MinGW before have to learn
about. This seems of dubious value relative to the gain enjoyed by the
few users who think this matters greatly, to be honest. The new format,
if implemented in replacement of the existing tar.gz archives, would
spawn a wave of complaints and demands for explanation and support. If
implemented in addition to the existing tar.gz format, it generates
extra work for the volunteer who prepares the release packages, who must
be absolutely sure that each archive contains exactly the same contents
as the other formats. How much that extra work matters depends on the
feelings of the volunteer and what kinds of demands are on her/his time
continuously. 

I suppose that if you want to see this implemented, one way would be to
become a volunteer yourself, and take on a role as a preparer of archive
packages for releases. You would need to talk to the project leaders
and, I imagine, establish yourself as somebody who could be trusted not
to muck things up. 

Another thing you could do is create a bit of documentation describing
and explaining all relevant phases of this new 7-zip tool, particularly,
I think, making a port to the MinGW build system (if it doesn't already
build on MinGW 'oob') available, so that people interested in working with
this new format could build their own tool. 

 Not speaking for MinGW "officially" in any way, just my personal
 opinions; 
   Soren A