> From: [...] Cees de Groot
> > I think the proposed roadmap misses out on a big 
> >opportunity.
> Fine. That's why we posted it, so you could make that clear.

Point.  Sorry if that came over a bit strong.

[... Lining up VM changes in 4.0...]
> The risk is probably that another round of breakage could be 
> introduced in 5.0. Is that bad?

For whom?  Many/most users shouldn't see a problem *except* that they
need to run an upgrade tool to mutate their particular long-lasting hunk
of bits to the new format.  If we can make that process straightforward,
format upgrades can be relatively painless.  If, however, we insist that
all users have to file out all changes, start with a new-format image,
and file in all changes, then I think we'll be in for a rough ride.

I suspect it would be possible to distribute a new-format VM + image
combination* that contains appropriate tools to work through a "3.x"
image and write a new image to disk that is in "4.x" format and could be
started and tested with the new VM.  This would allow would-be adopters
to test the new VM + image combination before committing to it, and
would be considerably more flexible than having to make the change from
within the existing image.  I accept that this would break the notion
that all changes can be distributed through the update stream; this
could be an issue.

		- Peter

* Why this way rather than going forward from 3.x?  Probably a personal
preference to demonstrate that the "4.x" format works for something :-).
It may be that going forward from "3.x" is a better bet, however.