[Bioc-devel] Attention Bioconductor Maintainers

But why not revamp the version number completely. Let
  x.y.z.w
be the version number. Let x.y correspond to the R version, let z
track the development and let w track release version changes. In that
way a typical package would be
  2.1.0.1
at the current R release. Changes to the fourth number would reflect bug
fixes, while development would result in an increase in the third
number. Once the R version bumps, everything starts from scratch.

Ok, this would break a lot of version tracking (that is a bit of an
understatement), but at least it would provide some version tracking in
the future, which is consistent with R.

Well, I am just throwing ideas around. Ignore them at will :) But
basically I can see the point in bumping version numbers when R/BioC
is bumped.

Kasper

On Sat, May 21, 2005 at 09:42:40AM +0200, A.J. Rossini wrote:

I think in this case I agree with Seth/Robert -- at the very least, we
probably should consider bumping the R requirements to track the
latest R (though NOT until near release).  It would be a complete
waste of time to backport core pieces of BioC back to the previous
version of R, for example, especially if dependent on core features
(or for example, using new features in the packaging files.  So there
should be minimal changes made, at the least.   Whether this merits
x.y.z -> x.(y+1).0, I think is more aesthetic than real.

best,
-tony



On 5/21/05, Robert Gentleman <rgentlem@fhcrc.org> wrote:

Hi Gordon,
    Do you have an alternative?
The problem, as I see it is the following:
   As R changes we track and use, and some times are the reason for,
those changes. So packages and developers need to do that on the devel
arm, it is going to morph into the next BioC release and be compatible
with the next release of R. If they do not bump their version numbers
quite substantially, then there are problems with version numbers
between release BioC, which must allow bug fixes, and hence must allow
version number increments, and those on the devel arm, which we hope
will have version number changes because people are active and adding
new features etc. I think it would be very confusing to start
interlacing these - but we are very happy to see anything that
developers want and that will help to provide for both a happier
developer experience and a happier user experience.

  This seemed to be the easiest way to achieve that, but certainly not
the only one - so please let us know what you think would work better.
If we can agree soon then we can implement it before too much divergence.


  Thanks
    Robert


Gordon K Smyth wrote:

Hi Seth,

I agree with you that the distinction between Bioc release and devel can be confusing and needs to
be clarified, but I don't think that this is the way to do it.  The trouble with incrementing all
the package numbers is that it breaks the basic principle that version numbers track changes in
packages.  It implies to a user that the packages have been changed, when they have not.  A user
will not be able to tell whether they should go to the trouble of converting to the devel version
or not.  It could even be considered to be a bit dishonest to give the impression that all the
packages have been updated and (by implication) improved when actually they haven't.

Anyway, this isn't a request for special treatment.  I think whatever you do you should do to all
the packages, and I'll work-around whatever.

Gordon

On Sat, May 21, 2005 9:53 am, Seth Falcon said:

Package version numbers have been "bumped".

In order to distinguish between the Bioconductor 1.6 release branch
and the devel version of Bioc packages, I have made the following
version number change to all packages included in the 1.6 release:

x.y.z  ==>  x.y+1.0

For example, Biobase was released at 1.5.12.  It's new devel version
is 1.6.0.  This way, patches on the release branch can continue to
increment the z number without any confusion with the devel version of
the package.

If you strongly object to this change in your package version number,
please contact me so that we can work out a compromise.

Thanks,

+ seth

--
best,
-tony

"Commit early,commit often, and commit in a repository from which we can easily
roll-back your mistakes" (AJR, 4Jan05).

A.J. Rossini
blindglobe@gmail.com

--
Kasper Daniel Hansen, Research Assistant
Department of Biostatistics, University of Copenhagen