[Mitgcm-support] CVS policy

[mitgcm-support at dev.mitgcm.org]

A,D,P,
 
  I have jotted down some notes on how we should be using CVS
Can you look at the notes and edit them as you see fit. Most
of the stuff is based on what we do already.
Notes are at 
 http://mitgcm.org/cgi-bin/cvsweb/cvs_tree_policy/policy.html

 I think the goal of the policies are at least three fold

 1. Help us coordinate things without too much effort
 2. Ensure that the product trees associated with our work don't
    get filled up with half-broken crap that nobody knows where
    it came from and what it does.
 3. Enable us to entrain some other people to maintain things.

 The thing I like about the current procedure is that the main CVS branch 
for mitgcm checkpointNN is always clean, always contains something that 
works. The checkpointNN branch does provide a fixed reference that everybody 
takes as their starting point.
Problems I think we have are

  1. Its quite time-consuming to work with the scheme we have.

  2. There is nowhere to dump things that are only partially finished.
     The development_tree/ area (see my notes) is supposed to help here. This 
     isn't perfect because there is no automatic link between development_tree/ 
     and checkpoints. However, at least it is better than how we work
     now i.e we variously
      a. check things in on the distribution branch without creating a checkpoint
      b. mislay things in a random directory structure
      c. create checkpoints with code which doesn't have full quality assurance i.e.
         no test case, highly-specific to one problem etc....

  3. After taking code from checkpointNN people can get themselves in
     a mess.

  4. The procedure isn't 100% defined! 

 What do you think? The great thing about CVS is we can even 
have multiple revisions of the CVS policy!! If we can agree 
on a common sense policy then we can figure out how to make it
happen. 

Chris