On Tue, Apr 22, 2003 at 03:13:45PM -0700, Bill Studenmund wrote:
> I think this scheme would work too, because at any one time the same
> version number would be in use for only one version at a time. Another
> option that has been floated was something like having current be 1.6.98.X
> for now, and then 1.6.99.X when we're in the 2.0 release cycle. Then 2.0
> comes out (from what was 1.6.99.final) and current's at 2.0.98.X.
I don't see the attraction of the even and odd numbering schemes, and
definately not when I was new too Unix (*BSD or Linux) trying to get
acquinted. I like the bumping to 1.7. directly after 1.6 release idea,
but I fail to see the friendliness in the even/odd stuff. There is
more in a numbering scheme than odd is development, even is stable.
What is the relationship between the odd and even versions? If there
is none, that is not obvious to new users either. I think that
relationship is very clear and simple in NetBSD, with as only
confusing thing the 1.6R < 1.6.1.

AFAICS, the Linux scheme has its merits, but only for a project
without project management using a central CVS tree that all developers
have access too (directly or indirectly).
I've been told (years ago) by a Linux developer that 
odd-number distributions were the main method of synchronization
between developers; he had to merge his code every time an
odd-distribution arrived and make a release of his own code for it.
He complained a lot about how much time that cost him.  

I think examplaining NetBSD's versioning and development as based
on a (CVS) source tree with tags and branches properly in
documentation is important. But I remember I got it explained the
first time I asked a NetBSD user. Likewise with
questions like: where do I send a question/bugreport/docfix? At that
time I asked similar questions about Linux to linux users. Still don't
know the answer to most of them for Linux. (Looking at the quality of
manpages on Linux, there either is no way to report problems there
or nobody knows about it!).

Feico.