Sarkela wrote:

> With a 
> couple of simple change sets, the most recent Squeak 3.1a can bootstrap a
> file based repository accessor from a file based repository.

???

> Revisions of changesets
> and configurations of them may be hand built into a repository. This allows
> both mainstream and experimental images to file in known combinations
> of change sets in known sequences from a well known place. I will publish
> the change sets for 3.1a this weekend with a message to the list when
> published.
> 
> The idea is that using the simplest possible capabilities in Squeak (file
> system
> access and filein) we can in fact build a system that will bootstrap whatever
> flavor of Squeak one needs. It is not modular (that is coming...)

I think this sounds great, even though I don't quite understand what it
does. :-) My worry with SWT was that it tried to change too many things at
once. As Paul mentioned, it is hard to make this work properly. It is also
much harder to 'release early, release often'.

I think we should try to modularize even the modularization effort itself in
this way, as much as possible. This sounds like a great step in that
direction. 

> Future development activities might include,
> incorporating the use of environments to isolate image based name conflicts
> an http based repository accessor (after all, the framework uses URLS to get
> revs)
> more unit tests to validate successful loading of configurations into
> particular images
> use ModSqueak to more carefully manage composition of images

Right, piece by piece is the way to go.

For example, one thing we need is to modify the compiler to direct the
resolution of all non-local (ie. UpperCase) names to a certain object
(representing a module/namespace). I don't know how far eg. Dan's
environments work got in this direction. This is a higly isolable piece to
work on. For now we only need to have one such object, for #Smalltalk, so
this code could go into the image right now without breaking anything.

Also, I think it is quite necessary for each of these pieces to be well
documented. I know this is to ask for the impossible with Squeak, but all
this "reading the image" is at best a nice fairy tale. Or else the blue book
wouldn't have sold a single copy, right?

What we'd really need is a Squeak Summit that could discuss and work out a
scheme to agree on, resulting in a design document. It is so hard to move
forward without a shared notion of where to g, and there are just some
things you can't do over the net. But I don't see how we could do that? A
meeting somewhere where there's a concentration of well-respected Squeakers?
Unfortunately that's not where I live... otoh, it seems there are a lot of
people between jobs in So Cal right now... another SqueakEnd perhaps?

just my two pieces of local currency,
Henrik