Meetings in Cambridge:

NOTE:  lines beginning with *** are unclear or not agreed to.  mostly disregard.

Administrative

	Biweekly international visits, starting May 18th ---> Montreal

	Bus Coupler two weeks late; use the four PCB's for prototype phase,
		but don't make more until bus coupler is complete.  It 
		should be debugged fully by May 12th.

	Bagley and Roth should maximize their interactions.  

	Serial IP project to Keith (~ 2 weeks)

	Coordination of modules: biweekly phone calls (interleaved with
		international trips)

	Within two weeks: Three Mac II's in Montreal, One in Lowell

Technical

	Full documentation needed in MTL; ADA must be contacted; Chris
	should be useful in following up on contacts at Apple for docu-
	mentation

	Mac OS uses 24 bits address, plus eight flag bits =32.

	currently have direct access, Video Cards <----> Disk, with 32-bit 
		addresses (for files only)

***	Screen Interface w/ Mac?  Should be direct. 
***		Lispm bitmap ---> Mac (no buffering scheme in memory)
***		No fakery inside the Mac needed to drive disk; no need to 
***		rewrite the driver, either.  Direct disk control of screen.

	Hard Limits: Max size of files? Max # of open files at one time?
		(for latter, it should be possible to increase the apparent
		limit through clever memory management on the Montreal side.

(the following idea should be considered, but it looks like files plus RWK's scheme
is better provided it can be big enuf)
***	Provide our own swapping space disk partitions (256 Meg)

	Provide UUCP for various intersite transactions within 3 weeks

	Tape backups: deal with them at the end of the project cycle.
		8 mm VCR tapes can store 2 Gigabytes, and would make
		for highly effective AAAI demoing

	Mac software needed from Montreal to convert complete LISPM pathnames
		into unambiguous Mac filenames. Use Mac resources to simulate
		(<pathname> :characters [T/NIL] :length-in-bytes (#) 
			    :byte-size (#) ... )
		although :characters and :byte-size seem redundant
		   
	Additions to Keith's document ("MAC-K-AREA.TEXT#4")
		Page 2: [HOW-MANY-K-MACHINES]
			Major-Command-Areas (One per extant K Machine)
		Page 3: Add *** RETURN-TO-MAC, *** RETURN-TO-K to list of 
		 	Major Commands

	Lambda microcode Sources to Montreal to facilitate first-pass interface designs.

	We might consider a facility for running a Lambda before the K has been completed. Replace
	kbd, mouse with Mac's [paging should be a last resort]; interface to
	communications area, staying away from microcode; use Lambda programs
	to build simulators.

	Cambridge will generate complete conditionalizations document(s).

	KBD I/O: 

		Need a single, fixed Buffer for mouse and kbd, with two pointers (I and O); 
		pointers = means buffer empty; Mac advances pointer upon
		encountering keystrokes.  USe the Mac File System char set as first-
		order approximation.
	        Montreal to propose full character set mapping.

	FILE I/O:

		Read-ahead; write-behind; setup-input-buffers scheme allows
		Mac to serve as feeder.  MTL should make a MAC-FAMILY-PATHNAME
		flavor mixin (patterned after the one for UNIX).  Scratch
		IOPB, replace it with a new block containing relevant info, 
		a la sys:cold;qcom.lisp, in which Request Blocks (RQB's)
		are used; command is in a leader, data follows. CCW =
		command and control word.  Mac does not have a multibus map; 
		therefore, build a gather/scatter memory scheme.  Swapping 
		should use RQB, block offsets.  LIMITS: Largest Mac file/
		Paging band/Paging file size.  Do not assume contiguous block
		file memory allocation; Use Mac files, and find out at boot
		time where all blocks are located for file, do a sort/merge
		on these blocks, leave files open, so we need deal only in
		physical addresses.

	MOUSE ISSUES:
          note stuff above under keyboard.
		Records size > word size; handling transitions vs. handling
		clicks; use a mouse-x and mouse-y, encode-kbd-clicks hack
		printed file UC-TRACK-MOUSE.LISP, in which it was observed
		single 32-bit register, 2 coords of 16 bits each, with three
		separate registers for the buttons; Mac has speed-dependent
		scaling as an option; Five char mouse message: buttons, dx1,
		dy1, dx2, dy2; each d being an 8-bit signed number; process
		is: figure out where you are in the five characters, simulate
		serial action of mouse, or simulate fake mouse using a click
		decoding scheme; if all are unchanged, do nothing.  [Start with
		the serial action scheme in MTL]  Tailor Mac's event manager
		to conform to MOUSE.LISP for clicks (already create events, 
		post them for LispM to read)

	CURSORS:

		Use the Lispm's. Make Mac's go away. LispM writes directly
		to the screen. 
		When in bit-array-in-memory-option, provide for K to 
		Tell Mac where to look for it using an interface
		to open- and set- blinker, and an array in memory. Copy just
		enough bits to make it work. Add mouse-x and mouse-y registers
		to Keith's scheme.

		Use One Buffer (synchronize mouse events)
		Interrupts vs. Lossage of characters when clicking
		

	TOOLBOX CALLS:

		Time: Use Lisp Machine's initialize-timebase, build an
		interface for the Mac to use this information. In general,
		use general purpose entry vector scheme (simple #'s) to
		gain access to all of them eventually with mainline calls.
		Linked-in Libraries, consisting of one vector plus an index
		to the right place. Do enough of this in MTL to ensure successful
		Mac calls for those that already work now.