Worthless updates

[Kevin Kofler]

Matt Domsch wrote:
> The _only_ reason to name something with a 'version' or a 'release' is
> to provide a set point for consistency, either in people's minds
> (marketing), or to provide a technical baseline for interoperability.
> If we continue to take the technical baselines, and move them ad-hoc,
> it's no longer a meaningful value.  We might as well move to SCM
> snapshot-of-the-day.

Did you read my reasoning about how the fact that we are currently a middle 
ground between such a rolling model and a frozen release is a great thing? I 
think you didn't get that point at all. :-(

The baseline we should be providing is: this is the stuff we ship now, our 
updates should only improve things, not regress or disrupt things. So e.g. 
pushing an upstream rewrite which is known to cause major feature 
regressions as an update is a no-no. But we shouldn't be so scared of 
regressions not to fix bugs or provide non-disruptive features!

Rawhide or Rawhide snapshots do NOT provide such a baseline. We need 
releases for that. But that kind of baseline is quite different (much less 
restrictive) from a version-locking baseline which you seem to be implying 
(even though you don't explicitly write that down anywhere).

        Kevin Kofler