[V3dot10] Re: Tentative of start 3.11

[goran at krampe.se]

Hi Edgar and all!

Btw, I am following this thread - but don't have time to comment much.

"Edgar J. De Cleene" <edgardec2001 at yahoo.com.ar> wrote:
> > Also, we should decide what we are going to put into the update
> > stream, or leave it alone for now. MC wasn't working, I doubt
> > installer scripts would be any better.
> 
> So I glad all know now same I know before 3.10 starts.
> And sad alternatives to change sets don't ready yet.
> I also have long talk with Colin and he promises finish MC2 and say MC2 is
> better suited.
> You said DeltaStreams don't works with Traits or I missunderstand some.
> It's DeltaSteams ready ?

No, not yet AFAIK. Matthew has been working on it, but I haven't had any
time since OOPSLA in October. Without having looked at Matthew's latest
snapshots I suspect we are approximately where we were then - the base
model is there and it works. Needs refinement, battle testing and UIs.

Regarding Traits - that is a darn good question. :) It should of course
support Traits - and I see no principal problem with doing that.

IMHO making a push on DS so that it at least can replace ChangeSets
without loss of functionality (we are there more or less I think) would
be VERY worthwhile. Out of the box (if we disregard all the funky stuff
we should be able to do with conflict detection etc) it should give us:

1. More information in the delta (=changeset). For example, not only the
new modified method but also how it looked before the modification.

3. A real model. A changeset is not "complete" - meaning that if you
load one changeset that changes one method, then load another changeset
that changes the same method - then if you inspect the first changeset
it will show you the latest code! Meaning that changesets don't actually
*hold* all their information. A Delta does this - for example, you can
load a Delta into the image but not actually apply it.

2. Ability to uninstall. This means you can apply 10 deltas, realize
that one in the middle broke things - then unapply all 10, remove the
faulty one, reapply the other 9. And other interesting use cases.

3. Atomic apply/unapply since we use Colin's SystemEditor - the new base
tool to actually perform code changes in the image atomically meant for
MC2. This means apply/unapply should be much more robust.

...and probably lots of other nice things. :)

[SNIP]
> Very glad Ralph, Pavel or Goran was in charge.
> I only wish help and learn.

In charge? I am not in charge. Interested and gladly involved - but I
don't have that much time to spare. I could try to concert a push around
DS though.

regards, Göran