[V3dot10] And to 3.10

Ralph Johnson johnson at cs.uiuc.edu
Mon Jan 22 12:40:15 UTC 2007


is very pretty.  It is extremely readable, and so people are more
likely to read it and to respond.  It is tied into Mantis, which will
encourage people to use Mantis.  This is good.

However, I see that page as a list of PROPOSED fixes, not what is
actually used to build the released image.  First, I think we want to
keep using Monticello like they did for 3.9.  So, "harvesting" changes
means converting the .cs files that are usually stored in Mantis into
changes to MC packages.  Second, the changes should all be tested on
our nonexistant test server.  Putting proposed changes in an Installer
script is a great way to let others look at them and try them out, and
will make it easy to test them.

People use "harvesting" to mean lots of different things.  it is much
more than just deciding which changes to put in the image.  That is
the fun part.  Harvesting is not fun, it is work.

Edgar, I don't have an opinion as to whether a particular set of
changes should be accepted or not.  I will firmly hold on to my lack
of opinion.  In general, we should accept changes when people want
them, all serious objections have been overcome, and they don't break
any tests.  Since Mantis is down, I can't look at it to see whether
there has been any discussion of these issues.  Has there been any?
Are people happy with the changes?  Did some senior Squeaker voice
approval?  If so, we'll take them.

ideally, each issue would be examined by a steward who would tell us
whether the change was good.  We'd still check that it didn't break
any tests, of course, but otherwise that would usually be good enough
for us.

But then the hard work of harvesting would begin!  Harvesting starts
with the decision to take the code, it doesn't end with it.

Keith, one problem with this way of doing things is that the scripts
make great documentation of what went into the image, but only if 1)
they are still there and 2) they haven't changed.  I am leery of
leaving importnat, long-term documentation on a wiki that I do not
host.  I think the solution is that when we harvest changes in a
script, we will save the HTML for the page on squeak.org as sort of a
diary of what we did.  Do you think that will work?


More information about the V3dot10 mailing list