Trustin Lee wrote:
Hi all,
The previous thread on versioning scheme is becoming very long, so I
want to
split it into two by summarizing what we agreed on. Alex summarized our
current versioning scheme here:
http://cwiki.apache.org/confluence/display/directory/Version+Numbering+Scheme
I agree with his idea mostly, but there are two points to raise:
1. 1.5 -> 2.0 vs 1.9 -> 2.0 vs 2.1 -> 2.2
1.5 is a very arbitrary number and this rule cannot be applied to the
future
similar changes. What happens if we move to Java 6? What happens we
already released 1.7 and if we move to Java 7? 1.9 has also a similar
problem; what happens if we already released 1.8? There's no way to
distinguish from previous releases because we are out of bullet.
Please don't put too much emphasis on the 1.0 -> 1.5 move as something
cute to mean that we went to JDK 5.0. We could have just picked 1.3 or
1.7.
The key point was we do not want to go to 1.1 because we're changing a lot.
2.1 -> 2.2 also has a problem that 2.0 will never be released, which is
kind
of weird. Releasing 2.0-M1/2/3/4...and then RC1/2/3... and finally 2.0 can
be a solution, but it makes the even/odd scheme pointless because we can
just use 'M{number}' to state that it's unstable. Of course, we can start
over with this whole new scheme.
2. Clarify the meaning of minor version number.
'may or may not' sounds very ambiguous. We need to clarify it.
You mean in my description of it on the cwiki?
We can just go 1.5 or 1.9 not settling down the version numbering scheme,
but we will encounter the same problem on and on. I'm not a perfectionist,
but I want to set up a nice rule that can be applied for as many cases
as we
can cover. Now the thread is hot, it's a great time to gather all opinions
and to improve our version numbering scheme.
I have to agree with Peter that this thread is in danger of turning into
a bike shed argument.
Alex
