* Philip Brown <ppb at usc.edu> [2007-10-24 20:42]:
> Lukas Rovensky wrote:
> > One more question from the maintenance point of view.  The pkg(1) makes 
> > possible to freeze a package at a specified version to stop update 
> > flows.  Currently this is decided by the user to use this constraint. 

> > Danek Duvall wrote:
> >> ......
> >> In addition, because of the "no dim sum patching" rule, ....
> 
> I would like to point out, that the capability, and the rule, mentioned 
> above (amusingly, in the same email), are to my viewpoint, directly in 
> conflict with each other.
> 
> The concept of allowing a user to freeze a package a particular
> version, is in essence, an anti-patch. (ie: instead of patching "up",
> it is patching "down")
> 
> To me, it is just another variant of "dim-sum patching"

  In its strictest form, a fully-specified constraint that includes the
  full version and timestamp could be seen as some kind of opposite to a
  patch.  Of course, it's not the same, as dependencies that require a
  version of the package newer than the constrained version will fail,
  and a system will end up on a version surface that lags the most
  recent updates.

  However, constraints needn't be so strict:  a constraint can say, keep
  me on this branch component, to this significance, and similarly for
  the version.  So a constraint allows me access to newer software, but
  along a path that I control (by limiting the scope of change).

  Since the constraints and dependencies are tied together, the
  combinatorics are significantly reduced from those available from the
  current library of patches.

  - Stephen

-- 
sch at sun.com  http://blogs.sun.com/sch/

Reply via email to