[V3dot10] [Q] About updates, take 3

[David T. Lewis]

On Wed, Jan 24, 2007 at 06:46:08PM +0000, Keith Hodges wrote:
> 
> >
> >You want people to update often, so stuff gets tested. Once someone 
> >did update, there is no going back with that image. Also, with a 
> >strictly additive stream of updates you can be much more sure what 
> >state the image is in.
> >
> >- Bert -
> >
> 
> I am not sure I agree, I have received many an update without having a 
> clue as to what it has done to my system. In contrast to using 
> monticello packages where I know what a package is supposed to do.

This was not my experience with the SqC update stream. Updates were
released frequently in groups of small change sets, each change set
having a suitable preamble. The preambles were summarized in a note
to the list so I knew in advance what I would get in the next batch
of updates. Overall I had a much better understanding of what was
entering the development image.

> 
> There is a whole lot of faith goes into pushing that update button.

Of course.

> 
> I think that monticello woud be much improved if you could attach 
> comments to a package after it has been committed to the repository. 
> This would provide more context to users as to what exactly they are 
> loading and how well it performs.

Yes.

> 
> Compare this with the dev image where you can review the script to see 
> exactly what has been loaded and how it has been configured.

Sounds fine. I just want to point out the value of the simple
update stream, because it may not be obvious to Squeakers who
did not experience the SqC release process (a few years ago now).
It seemed rather primitive to me at first, so it was one of those
"aha moments" when I realized how well it really works.

Thanks,
Dave