Re: Stable releases

[Neil Jerram]

Greg Troxel <address@hidden> writes:

> Right now we have stable releases at intervals best measured in years
> (0 in 2005, 3 in 2006).  While having them more often would be good,
> 2-3 days is way too ambitious.  I'd suggest as something that is
> achievable and would be useful:
>
>   release 2 months after last release if anything significant has
>   changed (where significant means new feature or bug fix)
>
>   release 6 months after last release if anything has changed at all
>
>   release after a 1 week cooling off period if a serious bug is fixed

That's fine, but 99% of the time I expect it to collapse in practice
to just the last point, because

- by definition, nothing should usually change in a stable series
  apart from bug fixes

- bug fixes usually happen in response to someone reporting a problem,
  and I find it difficult to imaging saying to that person "we've
  fixed your bug, but don't regard it as important and so will not be
  making a release until a couple of months' time".

> In my view, the main path to guile usage by other than the people on
> this list is via stale releases and then packaging systems.  This
> enables other people to choose to depend on guile.  Currently, that's
> a scary choice to make.

Not sure I understand.  What do you think it is that makes the choice
scary?

Regards,
     Neil