>>>>> On Fri, 4 May 2007 10:24:56 +0000, Julian Mehnle <[EMAIL PROTECTED]> said:

  > Andreas J. Koenig wrote:
 >> > If that's the case then I wonder why `r` ("reinstall
 >> > recommendations") in `cpan` on my system (CPAN.pm 1.76_02, Perl
 >> > 5.8.8) does NOT list Mail::SPF as being out of date, even though I
 >> > have the exact same version of Mail::SPF installed as the
 >> > complainant (2.004), and my CPAN.pm is even older than his.  Has
 >> > CPAN.pm's version comparison behavior changed since 1.76_02?
 >> 
 >> It has changed significantly between 1.7x and 1.8x. In 1.8x it started
 >> to treat multidot version strings as such.
 >> 
 >> The test 10version.t in the CPAN.pm distro compares how version.pm,
 >> Sort::Versions, Perl::Version, and CPAN::Version behave in comparison.
 >> I've just added two tests for the Mail::SPF case.

  > So does that mean that CPAN 1.9101's behavior (treating v2.004 and 2.004000 
  > as different) is correct and 1.7602's (apparently treating them as 
  > equivalent) isn't?

Unfortunately there's no consensus on what is correct.

I expect that that bahaviour that will be in version.pm at the time of
the release of perl 5.10 will set the standard behaviour and then we
all can hopefully switch to that behaviour.

  > I'm still wondering where the fault lies and what I should do to make the 
  > Mail::SPF distribution's version numbers (in META.yml, *.pm, etc.) work 
  > with older (1.7602) and newer (1.9101) versions of CPAN alike (optionally 
  > with newer ones only if that's a conflict).  What options do I have?

I think I answered this in my first mail in this thread:

  > Switching to pure numeric versions would help and switching to
  > versions with two dots but without trailing zeroes would also help
  > (I think).

After re-reading the thread I'd recommend switching to good old
strings that look like floating point numbers is still the most
reliable and beautiful solution. If you find multidot versions better
because Damian does or whatever, use them but better avoid trailing
zeroes there.

-- 
andreas

Reply via email to