Article 75871 of comp.arch:
Guy Harris (guy@netapp.com) wrote:
: (Followups to "comp.arch", although I'm not really sure *where* this
: belongs at this point....)

: Hugh Bonney <hbonney@netcom.com> wrote:
: >   Actually, Vanevar Bush's Memex machine paper describes what sounds like
: >   tiled windows on a screen, though I assume one was never built which is
: >   a patent requirement. That paper dates from the 1940's.  The title was 
: >   something like "As We May Think", but I don't remember - I've seen it 
: >   on a Web site somewhere.
: An AltaVista search for title:"As We May Think"
: turned up a pile of on-line versions, including one on the Atlantic
: Monthly Web site at
:     http://www.theatlantic.com/atlantic/atlweb/flashbks/computer/Bushf.htm
: (it having been published in the July 1945 issue).
: However, I didn't see, in a quick look, anything that sounds like tiled
: windows on a screen - it described microfilm storage of information,
: mainly, with display on a microfilm reader, and I didn't see anything
: that sounded like *multiple* displays.

	I think you are right. On reading it over, it has a bank of windows
	on a slanted desktop, so it's a series of real screens not 'tiled'
	ones on a single large screen. The effect is similar, but there are
	apparently controls to be associated with each one, not a single
	set with attention addressed at the machine level on a window at a
	time. (Top of pp 16 of the Atlantic Monthly version.)

	The part I remembered was the idea of creating links between the
	stored pages and having multiple links per page to thread them 
	together in multiple series that he called associative indexing. 
	Books, magazines, whatever, were to be delivered on film. The user 
	could generate threads also between any pages on the machine and 
	generate pages. The machine was to follow threads for you like 
	flipping pages, and there was to be a 'codebook' to pick up 
	discontinuous ones. The codes could be mnemonic for ease of use, but 
	there was to be a 'codebook' screen one could call up when necessary.
	The consideration for 'human factors' in all this is impressive - 
	it's supposed to be useful.

: (A fascinating paper to look at 52 years later....)

	To be sure. He also mentioned possibly storing numeric data on a
	little steel card by generalizing from the idea of wire recording.
	The encoding was to be a positional scheme of some sort. I hope that
	CS students get to read the Memex machine part, maybe to be told
	that it was written in 1945 afterwards. 
	
	The Memex machine would have been a good place for AI work to start 
	from. The idea of augmenting human abilities always was more useful 
	and possible than little autonomous gadgets that run around blowing 
	up the bad guys for you. (I don't know where this thread goes either,
	but it's not likely to be long; thank you for the information.)

	H.--