On Mon, Jun 02, 2003 at 04:41:05PM +0900, Takahiro Kambe wrote:
> In message <20030601.191737.46635560.uebayasi@pultek.co.jp>
> 	on Sun, 01 Jun 2003 19:17:37 +0900 (JST),
> 	Masao Uebayashi <uebayasi@pultek.co.jp> wrote:
> > > I found that previous canna package was 3.6.1 (== 3.6p1) and now isn't
> > > 3.6.3 but 3.6p3.  For consistency is it better to change 3.6.3?
> > 
> > This is a mistake.  It's It's okay to change back to 3.6.3.
> > 
> > (But I prefer '3.6p3' if this correction brings only small problems.)
> Is there any package using p[1-9]?  Most of them use digit only.

To quote from pkg_info(1):

             If the given pkg-name contains a shell meta character, it will be
             matched against all installed packages using fnmatch(3).  csh(1)
             style {,} alternates have also been implemented in addition to
             this.  Package version numbers can also be matched in a relation-
             al manner using the >=, <=, > and < operators.  For example,
             pkg_info -e 'name>=1.3' will match versions 1.3 and later of the
             name package.  The collating sequence of the various package ver-
             sion numbers is unusual, but strives to be consistent.  The magic
             string ``alpha'' equates to alpha version and sorts before a beta
             version.  The magic string ``beta'' equates to beta version and
             sorts before a release candidate.  The magic string ``rc''
             equates to release candidate and sorts before a release.  For ex-
             ample, name-1.3rc3 will sort before name-1.3 and after name-1.2.9
             Similarily name-1.3alpha2 will sort before name-1.3beta1 and they
             both sort before name-1.3rc1 In addition, alphabetic characters
             sort in the same place as their numeric counterparts, so that
             name-1.2e has the same sorting value as name-1.2.5 The magic
             string ``pl'' equates to a patch level and has the same value as
             a dot in the dewey-decimal ordering schemes.

i.e. why not use "3.6pl3"?

Regards,
Alistair