Re: packaging manual/ policy seem to *discourage* pristine source

[Philip Hands <phil@hands.com>]

Steve Greenland <stevegr@debian.org> writes:

> This is a bizarre interpretation. If it unpacks to the same code (such
> that "diff -r" produces no output), it's effectively the same source.
> Who cares about the packaging? (Yes, I understand that it screws
> up md5sums/whatevers on the archive. So what? The files should be
> md5summed, not the archive.)

Some people do, unfortunately.  Of course the example I'm going to
cite is a bizarre case, and puts the package (qmail) in non-free
anyway, but Dan Bernstein's distribution conditions read:

-----
You may distribute copies of qmail-1.00.tar.gz, with MD5 checksum d3033be700fd6f59ac0548c832652dd3. 

You may distribute copies of qmail-1.01.tar.gz, with MD5 checksum 1f606d6a5d1caaca6da6b6fa5db500bf. 

You may distribute copies of qmail-1.02.tar.gz, with MD5 checksum 01071fe52b5257adb4bb6bcf8149eb16. 

You may distribute copies of qmail-1.03.tar.gz, with MD5 checksum 622f65f982e380dbe86e6574f3abcb7c. 
-----

so re-packing or re-compressing is not an option in this case.

I imagine that a DFSG license that scrapes in on the basis of allowing
patches, could call for the main source to have a particular checksum.
Perhaps this is something we should keep in mind if we ever decide to
re-draft the DFSG.

Cheers, Phil.