>>>>> "Marcus" == Marcus Denker <denker at iam.unibe.ch> writes:

Marcus> submit changesets.

Marcus> Harvesting ist not about "we need to do X". It's just about
Marcus> "Someone has already done Y, let's get it in the release".

Marcus,

I understand your frustration. However, I think that moving Squeak
forward requires more cooperation from the harvesters themselves.

To take my own case, I had to revert to using a 3.7 image for my work
because of the refusal of the harvesters to guarantee that the 3.8
unstable stream will be incremental[1]. I really liked to fix bugs as
I found them instead of using workarounds. I really liked to distract
myself from serious work for a few dozens of minutes by going through
the BFAV and reviewing changesets. But I really *had to* go back to
3.7 because I don't feel like using two radically different images,
one for my serious work and one for my bugfixing/reviewing work, and I
can't loose time to reinstall all my environment in a new image (I
work on several projects at the same time) when the unstable update
stream gets modified back in time.

As I stated before, I think that Squeak, as other software projects,
will get better only if people can use the development version daily
and base their work on it, even if there are obvious stability
concerns in doing so (this is a development version after all and
people should be ready to get bogus updates and fix them).

Don't get me wrong: I am not complaining, I am just giving an example
where someone quite productive (regarding the rate of my submissions
and reviews at the end of 3.7 and the beginning of 3.8alpha) had to
stop because the current handling of the unstable stream[2] is not
compatible with the available resources.

I am still hoping that at some point you will change your mind, or
that you will add new harvesters if more are needed. The Debian
GNU/Linux distribution for example works that way: you can always go
forward, you never need to go backward; if something gets broken, it
gets fixed later.

  Sam

Notes: 
[1] Except of course if the update mechanism itself has been badly
    broken and needs to be fixed back in time.

[2] That is removing bogus updates rather than reverting them, because
    it takes much less time for the harvesters.
-- 
Samuel Tardieu -- sam at rfc1149.net -- http://www.rfc1149.net/sam