At 10:20 AM -0700 8/24/03, Jack Johnson wrote:
>Alan Kay wrote:
>>things could be carried in one hand). A number of different 
>>calculations from 1968 onwards indicated that about 1M pixels were 
>>needed to do "mostly everything". This number remained quite 
>>constant over the years and approaches (only being able to 
>>contemplate 500K pixels or less in the early days of PARC was one 
>>of the irritants that led to the overlapping window idea).
>
>Does this mean that earlier ideas (and potentially your ideal 
>environment) were tile-based, like Wirth's Oberon or Plan 9's acme 
>interface?

I'm not sure what you are asking here. There were a number of "iconic 
programming" schemes that were done in the 60s. Bert Sutherland's 
thesis in 1964 at Lincoln Labs was an iconically programmed multiple 
dataflow language (the first of these, I think). GRAIL was a 
flowchart language that was authored only using the tablet. AMBIT-G 
was a pattern matching language at Lincoln Labs (where the AMBIT was 
Carlos Christiansen, and the G was Paul Rovner). There were others.

>
>(Rob Pike attributes acme's UI to Oberon's UI, and Wirth says 
>Oberon's UI came from seeing *something* at PARC)

Wirth saw quite a bit at PARC, heh heh.

>
>Do you have mockups or screenshots of what you envisioned a 1M pixel 
>non-overlapping window environment would look like?

Oh I see what you are talking about. The calligraphic display era in 
the sixties didn't permit overlap of windows with opacity without 
doing so much work that no one including me thought of it.  So tiled 
interfaces of some kind were the norm. (This is an old idea.) 
Sketchpad III by Timothy Johnson in 1963 or so had a tiled interface. 
The Flex Machine that I did with Ed Cheadle in the late sixties had a 
nice calligraphic display with multiple windows and used a zooming 
idea as one of its display modes to deal with nonfocus windows (they 
would be around the periphery with the focus window as large as 
possible in the center). This idea has been invented numerous times. 
The best zooming interface (the first really good one also) in that 
time period was the Dataland UI of Nicholas Negroponte and Dick Bolt 
that developed into the Spatial Data Management System they did in 
the mid70s.

The computer science lab at PARC liked tiling nonoverlapping windows 
and the reduction of out of focus windows, and they stuck with this 
for a long time. Charles Simonyi took this and the Bravo editor (now 
known as Word) to MS when he became an early employee, and the first 
version of MS Windows was tiled (a really silly idea on a 640*480 
display).

I think Wirth attributes the tiled idea to PARC because this is the 
first time he had gotten interested in graphics and he was at PARC in 
the 70s and wasn't aware of the prior art.

>  How do you think it might have impacted the future course of events 
>if you had 1M pixels right out of the gate?

Not much, because my plan was to do 2.5D animated graphics at 
"Disney" frame rates from the gitgo, so quite a bit of thought was 
already going into the double buffering and blting schemes to make 
this possible on a 15MHz processor. Part of the overlapping window 
idea also just came because of this committment to 2.5 graphics. 
Having a background in drawing and painting and layout, etc., had me 
already thinking of "graphics" in "media" terms (in which things in 
layouts and collages do have to overlap, and which in animations also 
have to overlap). In fact the first movies of the early 1973 tests on 
the Alto actually showed text being continuously being dragged 
around. This turned out to be too slow by about *4, so we wound up 
doing the 10 fps 2.5 D animations as a kind  of special case rather 
than the general facility originally hoped for. This worked out fine, 
and there are quite a few movies that show what the Alto could do 
when aimed at animation: about 120 sq" of graphics (a bunch of Pegasi 
and 80 or so pingpong balls) at 10 fps.

Cheers,

Alan

>
>-Jack


--