>>>>> "Bert" == Bert Freudenberg <bert at impara.de> writes:

Bert> You should not be updating from the unstable stream, then. We
Bert> *will* modify it. For example, the best way to get a couple of
Bert> additions in that we missed when grabbing the Squeakland changes
Bert> is to just insert them in the right place a hundred updates
Bert> back.

Well, it depends of what you (harvesters) want. I use the unstable
stream because this way, while I do my development, I can easily test
the latest enhancements made to the image, review fixes and fix bugs
when I want to distract myself from "serious" work. Restarting from an
empty 3.8alpha and reloading the updates and reinstall the latest
version of the tools I use to fix bugs is likely to be much less
frequent if it is not in my work environment.

Moreover, after having been developing free software for about 13
years now, I strongly believe that the best software you can deliver
is the one you use yourself for your daily work.

Also, if I find a bug for which I can find a workaround easily while
working in a stable image, I am likely to use this workaround rather
than fixing the bug if I am in a hurry because it may have already
been fixed in an unstable update already that is not installed in my
image.

I understand that this may be easier to remove the update or to fix
the update stream back in time as of today, but this makes it very
hard for other people to follow it. I do not know the harvesting
process so I cannot comment on its effectiveness, but maybe it would
be a good idea to stop for a few days and concentrate on a way to ease
the harvester's hard job to allow them to create a new update from a
changeset in one click or so (doesn't that exist already?).

To make it clear: my request for successive unmodified updates is not
a rant, this is a plea for better development practices :)

  Sam
-- 
Samuel Tardieu -- sam at rfc1149.net -- http://www.rfc1149.net/sam