[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Proposal: Debian release numbers

Eduard Bloch wrote:

> Just my 0.02EUR: Let's call it after the year when it is going to
> be released. For example, if Sarge will be ready in this year (what I
> doubt), it would be called 3.1, and Sarge+1, released 2004, will be
> Debian 4.0, Sarge+2 somewhere end of 2002 would be called 4.1, etc.

I think you mean "Sarge+2 somewhere end of 2004", right?

And why would Sarge+1 be called 3.1? 3.0 came out last year, so if I
understand your proposal, Sarge+1 will be called 4.0 because it will be
the first release this year (or next).

> Why this? A long development period is an indicator for major changes,
> eg. gcc-3.2 transition.

However, long development periods don't necessarily correspond to
changes of years as nicely as you might like. Let's say we put out Sarge
in January 2004, then Sarge+1 in November 2004, then Sarge+2 in March
2005 (I know this isn't likely given Debian's history, but consider it
anyway). By your logic and this schedule, Sarge will be 4.0, Sarge+1
will be 4.1, and Sarge+2 will be 5.0. However, note that the time delta
between Sarge and Sarge+1 is 10 months, while the delta between Sarge+1
and Sarge+2 is only 4 months. So why does Sarge+2 get a new major
version while Sarge+1 doesn't?

Another point is that by your proposal, every new Debian release will
get a new major version, since Debian, historically, tends to put out no
more than one new release per year, if even that. So in practice, your
proposal is equivalent to "just use a single integer version number".

> Just changing the major number to be proud of
> some new version does not work, since the freeze time is so long that
> that we often release with outdated software. Numbers and names are for
> marketing people.

This is largely true. It doesn't really matter what we call our releases
as long as there's some clear indication that it's a new release. Any
number greater than that of the last release will do, when you get down
to it. So I prefer the idea of simplifying the number to a plain integer,
as suggested at the start of this thread. It's simple, it's understandable,
and it will put an end to all the silly arguments we have EVERY RELEASE
about whether the new release should be N.[1-9] or (N+1).0.


Attachment: pgplJdIGFlRjo.pgp
Description: PGP signature

Reply to: