undo (was: Scrollbar morphems?

["Jecel Mattos de Assumpcao Jr." <jecel@lsi.usp.br>]

On Wed, 6 Sep 95 15:59 PDT
lpd@aladdin.com (L. Peter Deutsch) wrote:
> > I too argued against accept/cancel.
> > I think that an "accept" should happen after each keystroke.
> > What you see should always be what is real.
> 
> I think this is only a good idea if three things are provided:
> 
> 	- A cheap and simple way to create checkpoints, so that if you mess
> up, you can get back to a state of your choosing.

Yes - when I started my project of a "commercial Smalltalk machine" so
many years ago I knew that I would have to replace the snapshot/change file
system. I was good enough for lab use, just as the old 5 1/4 inch floppies
were good enough, but to have real people use it we would need something
better.

> 	- Fine-grained forward and backward (undo) replay of everything you
> did, step by step.

Anyone who has designed a chip using Magic would agree with you 100% ( this
program lets you undo a complex chip design all the way back to a blank
screen and then redo all the way to where you were initially!! )

> 	- An implementation of incremental accept that adds no perceptible
> delay to activities that don't otherwise have such delays.

You would have to add automatic "transactions" to the user interface,
but I already have to do that anyway simply to have a meaningful
interpretation of control-C.

> In my opinion, if you don't have those things, it destroys the fearless and
> effortless quality of investigation and evolution that computers should be
> able to support.

How true! When you change the color of a complex morph and get very
different results from what you expected, it is a very painful
experience to discover how much work you are going to have to get
things back to the way they were. You will be very afraid of trying
different things from then on...

-- Jecel