perm filename JBR.MSG[MPX,SYS] blob sn#699494 filedate 1976-06-06 generic text, type C, neo UTF8
COMMENT ⊗   VALID 00029 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00004 00002	∂23-OCT-75  0118	1,RWG at TTY41  0118
C00005 00003	RAID, FRAID
C00008 00004	PALX
C00010 00005	PDP-11, PDP11, DDT
C00011 00006	KL10, KL-10, MICROCODE
C00015 00007	AUGAT
C00024 00008	∂14-MAR-76  0915	FTP:RMS at MIT-AI   
C00028 00009	∂22-MAR-76  1712	RH   at TTY115  1712	tv-edit  
C00032 00010	things to do:
C00033 00011	∂12-APR-76  1130	JMC  	Comparison of KA and KL 
C00035 00012	∂13-APR-76  0618	FTP:JM at MIT-MC    
C00039 00013	TELNET
C00041 00014	∂16-APR-76  1217	JBR  
C00042 00015	∂21-APR-76  0857	TVR  	New XTHICK text offset in D  
C00043 00016	KLDCP
C00044 00017	Datamedia Simulator
C00045 00018	∂28-APR-76  1513	RH   
C00046 00019	∂01-MAY-76  0523	JBR  
C00047 00020	∂11-MAY-76  1358	RF  	11DDT manual   
C00048 00021	∂16-MAY-76  0302	JBR  
C00049 00022	∂21-MAY-76  1430	RH   
C00050 00023	∂27-MAY-76  0430	100  : ME 
C00051 00024	∂29-MAY-76  1915	REG  
C00052 00025	∂30-MAY-76  0334	JBR  
C00053 00026	Network Problems
C00054 00027	∂04-JUN-76  1751	FTP:MOON at MIT-MC	 KL10 parity errors  
C00057 00028	∂06-JUN-76  0246	FTP:David A. Moon  (MOON @ MIT-MC)	 FM Parity Mod 
C00058 00029	∂06-JUN-76  1434	FTP:Jack Holloway  (H @ MIT-AI)    
C00059 ENDMK
C⊗;
∂23-OCT-75  0118	1,RWG at TTY41  0118
1.0e38=377454,,732312 (41...) or 377454732313 optimal



RAID, FRAID

∂20-DEC-75  0046	REG  	FRAID    
1. Is there a way to zero a block of "core" in FRAID?
2. When examining a 2 segment file, you should prohibit looking
   above C(HILOC) in the lower, as that is somewhat misleading.




∂11-DEC-75  1649	JBR  
look at opdef mumbl[0] wrt RAID



FRAID is confused: <ctrl><meta>I followed by <meta><top>D gives many pretty
lines of "FRAID                 FRAID"



FRAID doesn't allow spaces in middle of file name, and doesn't know
↓file↓.ext sysn// syntax.



fix IN undefined bug in RAID.



	When either JRST DDTMUMBLE or R FRAID causes RAID to be
entered the first time, when on tty, it prints the header line as it should,
then the first thing you try to do it prints the header again before
doing what you said.  I believe the first-time flag is not getting cleared
properly at startup header-printout.
	A similar bug of the opposite sense happens if you had done some
stuff in RAID and did EXIT$$$Y followed by SAVE <FILENAME>, then later you
GET <FILENAME> and DD.  In this case it prints absolutely nothing to 
indicate that you successfully entered RAID.  Also, if your program does
JRST DDTMUMBLE, absolutely nothing is printed because it has the first
time flag still cleared from before the core image was ever saved long ago.
I suspect this second bug can't be fixed in a TOPS-10 system, except possibly
by a special check for EXIT$$$Y to turn back on first-time-flag then.



RAID thinks that IN is unknown when typed in, but will display IN for
056000,,0.  This happens for both .R RAID and .R LOADER ↔/V$ .DD↔.

PALX

∂08-JAN-76  0927	RF   
To:   MJC, JBR    
COM FOO←BAR.PAL produces BAR.BIN instead of FOO.BIN

∂09-JAN-76  1143	MJC  	SNAIL feature 
To:   RF
CC:   JBR    
The left arrow specifies creation of a DMP file, not a REL file.
COM FOO=BAR.PAL would have done what you wanted.  Probably I should
turn "←" into "=" in this context (I already do it for the PUB and
POX commands).
-----------

∂12-FEB-76  1308	RF  	PALX 

This is the current state of PALX:

The old favorite:
    PALX.231[1,JBR]
    PALX.DMP[1,3]

The new version capable of handling (I think) extended addressing:
    PALX.232[11,SYS]
    PALX.DMP[11,SYS]

The differences twixt the two:
    PALX.DIF[11,SYS]

The current flaws in PALX:

    1) If I say COMPILE FOO.PAL[BAZ,GRP], and I am not aliased there,
    PALX says it cannot enter the file.  If, however, I R PALX and
    say FOO←FOO[BAZ,GRP] it has no trouble.

    2) If I am doing the above (that is, running PALX on
    FOO[BAZ,GRP]), if FOO has an .INSRT to a file, PALX will take it
    from [BAZ,GRP], even if I am not aliased there.

My requests:

    Can you look into these PPN bugs, and maybe fix them?
    Can you suggest where we should save which of the various copies?

My thanks.
	

PDP-11, PDP11, DDT

∂25-NOV-75  1805	1,JBR
LOOK AT USER STACK CHECKER IN DDT



∂11-DEC-75  1027	100  : TED	11 DDT   
WHEN YOU TYPE THE CONSTRUCT A<B>C$Z  A CONTAINS C, AND A+1 TO B
INCLUSIVE CONTAIN 0.  I BELEIVE ALL LOCATIONS AR SUPPOSED TO BE
CLEARED TO C.




∂12-DEC-75  1455	RF  	10-11 interface
When the interface is given an address like 160000, does
it assert the upper two bits to put it into I/O space
at 760000?  If so, is this feature resettable?  If not,
is this feature settable?




KL10, KL-10, MICROCODE

∂03-FEB-76  0736	FTP:RG at MIT-AI    
Date: 3 FEB 1976 1034-EST
From: RG at MIT-AI
To: JBR at SU-AI

HI.  RE UCODE, SURE.  MOON IS THE KEEPER
OF THAT THESE DAYS (OR AT LEAST HE KNOWS THE CURRENT STATE),
SO I WILL TALK TO HIM AS SOON AS POSSIBLE.  A COUPLE OF POINTS:
  1)  OUR MOST CURRENT MICROCODE IS "REVISION 7" MICROCODE.
(OUR MACHINE IS CURRENTLY AT REVISION 6, APPARENTLY THE LAST KL-10
AT REVISION 6).  YOUR MACHINE IS ALMOST CERTAINLY AT "REVISION 8".
THIS ONLY MAKES A FEW FAIRLY FINE POINTS WORTH OF DIFFERENCE, HOWEVER.
  2)  THERE ARE A NUMBER OF ASSEMBLY SWITCHES CONNECTED WITH
WHAT PAGING MODE TO USE, SOME OURS AND SOME DEC'S.
     A) KI MODE (HARDWARE PAGE-TABLE REFILL, USED BY DEC 10/50)
     B) KL MODE (DEC HACK NOT USED BY ANYBODY.  THIS SORT OF IMPLEMENTS
        A REVISED VERSION OF THE BBN PAGER.  NOTE: IT IS OF PRIOR
        VINTAGE TO DEC S "CURRENT" EXTENDED ADDRESS MODE SCHEME.
     C) ITS MODE.  WHAT WE USE.
  NONE OF THESE IS THE NEW DEC EXTENDED ADDRESS SCHEME, WHICH BY THE WAS,
HAS BEEN CLAIMED TO BE WORKING HARDWARILY.
  3) IN ADDITION TO THE PAGER ITSELF, YOU ARE PROBABLY GOING TO WANT TO
WORRY ABOUT THINGS CALLED THE EPT (EXEC PAGE TABLE) AND UPT (USER PAGE TABLE).
THESE THINGS ARE USED FAIRLY EXTENSIVELY FOR VARIOUS PURPOSES NOT
CONNECTED DIRECTLY WITH PAGING, SO YOU WANT TO BE AWARE OF THAT
WHEN LAYING OUT YOUR DATA STRUCTURES.
  YES CHEOPS SEEMS TO BE A SUPER WINNER.  "LAST BUG" (WHICH WAS SOFTWARE)
WAS FIXED OVER THE WEEKEND.  HAVENT REALLY PLAYED TOO MUCH WITH IT YET.
  JUST TRYING TO GET STARTED AGAIN ON LISP MACHINE SOFTWARE.  ITS IN ABOUT
THE SAME STATE AS SEVERAL MONTHS AGO (IT HAS BEEN AGING PEACEFULLY
WHILE I WAS HACKING CHESS MACHINE).  HOWEVER, WE (TK AND H) DID GET
THE OLD 16K AMPEX CORE MEMORY HOOKED UP TO IT A FEW DAYS AGO, AND IT
SEEMS TO WORK.
  SEE YOU LATER, AND WILL GET BACK TO U SOON ON UCODE.
-------

∂19-FEB-76  0115	FTP:MOON at MIT-MC  
Date: 19 FEB 1976 0411-EST
From: MOON at MIT-MC
To: JBR at SU-AI

WE RECENTLY RECEIVED A COPY OF DEC EK-KL10A-BD-001,
"KL10 FIELD MAINTENANCE PRINT SET SUPPLEMENT."  IT'S
ABOUT TWENTY PAGES OF PRETTY PICTURES OF DATA PATHS, ETC.
YOU MIGHT WANT TO GET ONE.
-------

∂26-FEB-76  0837	LES  	KL10 delivery 
To:   JMC, TED, REG, JBR    
Just talked to Stafford who said he was told our KL10 will be shipped
on March 26.  He is awaiting TWX confirmation and will give it to us
in writing as soon as he gets it.

AUGAT

						M.I.T. Artificial Intelligence Lab
						545 Technology Square
						Cambridge, Mass.  02139
						253-7807
						February 3,1976

Len Doucet
Product Manager, High Speed Packaging
Augat Inc.
33 Perry Ave.
Attleboro, Mass. 02703

Dear Mr. Doucet:

	M.I.T. is interested in a quotation for setup and production of a special
purpose wirewrap card similar to the existing 8136-PG-21-180-TG with the following
exceptions:

1)  Edge connectors having 34 instead of 26 pins. (can we fit 40 pin connectors?)

2)  Lower half of each connector soldered to the ground plane.

3)  Ground and power planes extended to the edges of the board (cutout around
    the mounting screws)

4)  Depending on the costs involved, ground plane interconnects between each pin of
    the pattern, as in the newer universal cards

5)  an extra row of power and ground pins just below the lowest row of socket
    patterns, as if a new row of sockets was about to start.

6)  Power and ground pins with socket type pins rather than the flat-headed pins

7)  Silkscreened rather than etched identification of the sockets and pin numbers

       We are interested in making this our "standard" wire-wrap panel and would
like to see it as part of the Augat standard product line.  If there are
particular features of the board which make it excessively difficult or costly
to manufacture, I would appreciate it if you could give me that information so
that compromises might be made where they are most effective.

       In examining our latest catalog I also note that Augat has not yet
announced an adapter for mounting 24 and 28 pin dips on 16 pin pattern boards,
which occupies only three slots.  I know of the version which mounts the dip, up
to 40 pins, vertically, and takes 4 slots, but we have been using a version
internally which mounts up to a 28 pin dip in 3 slots.  I believe it would also
make a good product.  Another idea along the same lines is an adapter for two 20
pin dips into three 16 pin sockets.  I know that we, at least, would find that a
useful product.


						Sincerely,



						Tom Knight


Phone call 2/19/76

Augat part # AX35-76-1G1

1-5	$387.75
5-9	$346.50
10-24	$322.75

Delivery 4 weeks if youpress him, 6-8 weeks if not.
-----------

∂18-FEB-76  0353	TAG  
THAT AUGAT PANEL IS NEARLY IDENTICAL TO ONE I ASKED LENNY ABOUT MAKING
AT WESCON. WHY DOES MIT INSIST ON LOSING SEPERATE POWER AND GROUND PINS?
99.998276% OF THE DIPS ONE USES HAVE POWER AND GROUND ON THE CORNERS. WE SHOULD
TRY TO CONVINCE THEM THAT DEDICATED (PG20) STYLE P & G PINS ARE THE WAY TO GO,
AND THEN SHARE IN THE DEVELOPMENT COSTS. LENNY TOLD ME IT WOULD COST
LESS THAN $500 TO TOOL UP FOR SUCH A BOARD. BY THE WAY, THE GROUND THREADING
BETWEEN PINS IS UNNECESSARY ON A PATTERN BOARD.
-----------

∂19-FEB-76  1403	FTP:TK at MIT-AI    
Date: 19 FEB 1976 1701-EST
From: TK at MIT-AI
To: jbr at SU-AI

I got a response from Len Doucet at Augat with a part number and
prices, which I tacked onto the end of the letter, so if you are
interested, its there.  The quote is for 40 pin i/o connectors.
-------

∂14-MAR-76  0915	FTP:RMS at MIT-AI   
Date: 14 MAR 1976 1213-EST
From: RMS at MIT-AI
To: STAN.K at MIT-AI, JBR at SU-AI, REG at SU-AI

THE DEC VERSION OF @ FINALLY WORKS.  THE CURRENT SOURCE
IS GLS;@ >, WHICH ASSEMBLES OK FOR ITS AT LEAST.
TO USE IT, YOU NEED TO WRITE CODE TO
READ AND PRINT YOUR FLAVOR OF PPN, AND
ANALYSE YOUR FLAVOR OF FONT FILE (TO GET CHARACTER SIZES OUT).
-------

∂22-MAR-76  1712	RH   at TTY115  1712	tv-edit  
hi, my name is ronny, i am a new student working for cordell green and
have a question to ask you about tv-edit on the imsss system.
what terminal do you tell you are when using tv-edit on a data-disk?
my second question is how do you connect to the AIsystem from IMSSS
given you have an account on both systems? thanks rh (ppn)


∂23-MAR-76  1212	RH   at TTY116  1212	datamedia    
hi, thanks for your answers, the assignment of a datamedia to this
datadisk is not succesfull for the tv-editor at imsss though. neither
is the assignment of an imlac, so maybe you cannot edit from uphere?
do you know if there is a way of transfering files from imsss to this
system?
I sometimes dialup from home with a datamedia, later it may become
a TEC unfortunately (or is that not a big problem)
I don't have an office (yet?) but work in the annex, I will drop in
your office one of these days, thanks, ronny

∂31-MAR-76  1428	RH   
HOW DO I MAKE A PAGE MARK? RONNY

You make a page mark by saying control-X then M<CR> (M stands for mark).
This puts a page mark at the cursor and leaves you looking at the new
page on the top line.  You can delete the page mark at the bottom of
a page by saying control-X then DEL.  --jeff

∂01-APR-76  1909	RH   via IMSSSS
thank you i did not understand that a mark is a pagemark

Ah, but a mark is NOT a pagemark.  A mark is a place marker which
is used to remember certain places in the file for quick and repeated
reference.  They go away when you exit the editor.  Page marks delimit
pages and are stored in the file, so they remain when you exit the
editor.

 set a mark with	αβM
 remove a mark with	αβ-αβM
 move cursor forward
   to next mark with	αM
 move cursor backward
   to next mark with	αβ-αM
 remove all marks with	αβ0αβM

 set a page mark at
   the cursor with	αβXMARK<cr> or αXM<cr>
 delete the page mark
   at the bottom of
   this page with	αβXDELETE<cr>

				--jeff
∂13-APR-76  1118	RH   at TTY123  1118 via IMSSSS	NAN
HI, I JUST READ YOUR MESAGES ON PAGEMARKS AMD MESSAGE-EDITING, THANKS,
ALSO LETS GO TO NAN NEXT MONDAY, IS THATOK WITH YOU? YOU MAY NOT
RECOGNIZE ME THOUGH, BECAUSE I CUT MY HAIR, BYE RONNY

things to do:

try to borrow LA501 from Tektronix.  Also get prices.

@ IS NOW SNARFABLE.  YOU NEED ONLY MAKE IT UNDERSTAND
YOUR FLAVOR OF FONT FILE (SEE FNTCPT) AND YOUR FLAVOR OF PPN
(SEE FPARSE, 1INSRT, FLOSE0, FILOUT).

call Ramlor about pc cards

∂12-APR-76  1130	JMC  	Comparison of KA and KL 
To:   DCO, LES, RWW, JAM, TOB, REG, JBR    
Please time some typical or large jobs on the present system,
and can them so they can be run again when the KL comes up.
Our planning depends on knowing how much faster the KL-10 is.

∂21-APR-76  1359	LES  	Comparison of KA and KL-10   
To:   DCO, RWW, JAM, TOB, REG, JBR    

In making timing runs to compare KA and KL performance, you should
note that CPU times in our current system include a substantial
amount of overhead, so that measured CPU times on identical runs may
vary as much as 50%.  In making comparisons, you should take several
runs on each machine under similar load conditions and compare on a
consistent basis (e.g.  mean with mean or min with min).

∂13-APR-76  0618	FTP:JM at MIT-MC    
Date: 13 APR 1976 0919-EST
From: JM at MIT-MC
To: jbr at SU-AI

i don't wish you to trnasfer macsyma to isi or anywhere else.
our current research funding is based entirely on people 
paying for the use of the kl-10 here. if the system were
available elsewhere, we would potentially and frankly actually
lose a significant amount of our funding. 
maclisp,on the other hand, is quite a different story. we are
very interested in bringing it up at isi and potentially other
tenex sites. the reason is that we would like to be able to run
our programs(martin's natural language stuff and the medical 
diagnosis stuff) on tenex sites.
-------

∂16-APR-76  2329	EK   via AMET  
Too bad! I wonder what Moses has against TENEX.  Just found out that I have
to have a paper done within the next week or so, so I guess I'll let
macsyma slide for a while.  Dick said he thought we (SAIL) might
be getting a version to put up here (McCarthy wants it), so maybe that
will help some.  Anyway, guess I'll let it go for a while.  Thanks for
trying anyway, and let me know if you hear about
any new developments.
Elaine

I got a note from Moses saying that they didn't want it up anywhere else
since they depend upon revenue from MACSYMA users for their support and
he didn't want them to be able to get it anywhere else.  As to putting
it up here, there is no way that there could be a timesharing system here
that would allow a job that big in the next six months.  What do you want
to do about all the files at ISID (mostly *.FAS and MACLSP.* and MACLISP.*)?
You might conceivably be able to convince Moses that any version you might
put up at ISI would be for your own personal use and not for general consumption.
Lastly, there is the problem that MACSYMA is too big for the two segment
MACLISP that runs on the DECSYSTEM and there doesn't seem to be any enthusiasm
at MIT for fixing that.  They say they'd rather work on the real TENEX version.
That, however, is also not likely to happen too soon.  Sigh, it's all a great
conspiracy.  Well, good luck on your paper.   --jeff
TELNET

∂13-APR-76  1812	FTP:REM at MIT-MC   
Date: 13 APR 1976 2008-EST
From: REM at MIT-MC
To: JBR at SU-AI

Because of the frequent occurrances of MAIL getting wedged under a telnet
server and TIPs dropping carrier etc., I think it would be a good idea
to DETach rather than KILL jobs under a telnet server that dies or
self-destructs.  If you're worried about accumulation of detached jobs,
make them subject to autologout after 1/2 hour if nobody has come
back to attach to them by then.  Of course, 100,100 temporary jobs could
be killed immediately to protect against unauthorized users constipating
the system.  MIT-ML,MC,AI detach rather than kill (except temporary jobs
not really logged in) and this proves to be a great winner almost always.
-------

∂16-APR-76  1217	JBR  
fix midas bug in file name reader in .insrt (nsf, read what instead)

∂16-APR-76  1219	JBR  
also fix no such file PPN typer.

∂18-APR-76  0ND X Y))

			  ((ATOM X) (CONS X Y))

			  (T(CONS (FLAT1 (CAR X) (FLAT1 (CDR X) Y)))))))

∂27-MAY-76  0430	100  : ME 
Norm wants you to call him if you are going to the softball party Mon.
Try ext. 7-6051.  I have a map to his house.

T:&($  Could you please fix αK so that re-extends
lines which had OFFRIGHTed text on them?  --- Tovar

KLDCP

∂21-APR-76  1316	JBR  
BUG IN KLDCP WHEREBY TENSP DOESN'T WAIT LONG ENOUGH IF MACHINE
IS DOING NXM TIMEOUT.

Datamedia Simulator

∂22-APR-76  1516	BPM  	Datamedia simulator
It doesn't seem to know about CONTROL-Y for setting tabs.

You want to be able to simulate our Datamedias exactly so that you can
debug DM programs at any Datadisc--eventually even the one DM that ME has will
go away to ALS's house.
∂28-APR-76  1513	RH   
lets go to the varsity not next week but the week after that monday
or tuesday or wednesday? ok, maybe i can this coming week but i dont
think so. ronny

∂01-MAY-76  0523	JBR  
HACK UP A DIAGRAM OF CRAM BITS IN EBUS FORMAT

LOOK FOR PLACES IN MIT UCODE WHERE VMA IS ON DIFFERENT LINE FROM LOAD AR

∂11-MAY-76  1358	RF  	11DDT manual   
The manual you wrote was updated somewhat for the Hand-Eye progress
report of last year.  I just took that update, added some new stuff
for the $U and $$U instructions, and made a file called 11DDT.POX[HAL,HE].
I think this should become the standard documentation.  Would you look
at it and tell me what renaming, if any, is in order?

∂16-MAY-76  0302	JBR  
MODIFICATION TO PMP CHANNEL FOR MPX:
CONNECT CARRY OUT OF WC CARRY GENERATOR TO SUPPRESS MAPPING ON MEMORY
READ CYCLES.

∂21-MAY-76  1430	RH   
HERE IS MY FINAL SOLUTION WHICH WORKS, I DONOT THINK YOUR DOES...?

DEFINEQ((FLAT (X)(FLAT1 X NIL]

DEFINEQ((FLAT1 (X Y)(COND ((NULL X) (APPECH A BOARD. BY THE WAY, THE GROUND THREADING
BETWEEN PINS IS UNNECESSARY ON A PATTERN BOARD.
-----------

∂19-FEB-76  1403	FTP:TK at MIT-AI    
Date: 19 FEB 1976 1701-EST
From: TK at MIT-AI
To: jbr at SU-AI

I got a response from Len Doucet at Augat with a part number and
prices, which I tacked onto the end of the letter, so if you are
interested, its there.  The quote is for 40 pin i/o connectors.
-------

∂29-MAY-76  1915	REG  
3. I think you put a bug in KLDCP.  When I did V SUT.RAM, it said "BOOT TOO LONG"
and I had to RI to make V work.

4. I wrote something to handle address breaks.  When first I started this afternoon
(using an old SYSTEM.DMP) there was this bug where someone was stuffing shit
into (one word) of the dpyser clock queue.  So I tried using address break to
find it, and it went into hiding. 
∂30-MAY-76  0334	JBR  
CHECK PC THAT IS STORED ON ILL MEM REF DUE TO FALLING OFF LAST PAGE.

Network Problems

∂31-MAY-76  2111	FTP:MOON at MIT-MC	NETWORK LIAISON BUSINESS  
Date: 1 JUN 1976 0002-EST
From: MOON at MIT-MC
Subject: NETWORK LIAISON BUSINESS
To: JBR at SU-AI, JAF at SU-AI

DO YOU KNOW THAT YOUR SYSTEM SENDS SPURIOUS "RRP" MESSAGES
WHEN IT COMES UP?  MAYBE IT THINKS IT'S SENDING "RST".
OR MAYBE OUR SYSTEM ONLY THINKS THEY'RE SPURIOUS?  BUT WE
GET THEM FROM NO ONE ELSE.
-------

∂04-JUN-76  1751	FTP:MOON at MIT-MC	 KL10 parity errors  
Date: 4 JUN 1976 2050-EST
From: MOON at MIT-MC
Subject:  KL10 parity errors
To: JBR at SU-AI

Sorry not to reply sooner, I haven't read my mail in a few days.

We never tracked down the ARX parity errors to their source.
I guess DEC never did either.  Glad you found that one,
maybe they'll send out a fix for it someday. I noticed
revision 7 had a good deal of screwing around with the
timing of those 69 different page fail signals.  Haven't
seen any later revisions yet.

We mostly found that fast memory parity errors were getting
caused by writing into fast memory in the same instruction
as NXT INSTR - for instance, see STAC1: in our version of
ucode, we had to make it J/FINI instead of NXT INSTR.
Before that change JFFO would get fm parity errors one in a
thousand times or something like that.  In all cases where
we checked, the data was always right and the parity was
wrong.  The FM parity generating logic is totally screwed
up.  (Have you looked at it?!) For some reason, although we
had to change STAC1 we didn't have to change STAC, only
difference is AC1←AR vs AC0←AR. We also had a AC0←AR,B DISP
which tended to write bad parity in fast memory.
Putting TIME/3T instead of 2T made it happen MORE often.
∨
-------

∂06-JUN-76  0246	FTP:David A. Moon  (MOON @ MIT-MC)	 FM Parity Mod 
Date: 6 JUN 1976 0543-EST
From: David A. Moon  (MOON @ MIT-MC)
Subject:  FM Parity Mod
To: JBR at SU-AI

Well, send me a description of it and I'll think about it.
We haven't had one in a long time, but maybe with it we could
take out various micro code kludges to prevent them.

I'm not sure if we'll want to make it / be confident that it
will work in our machine, but when I see it I'll know.
-------

∂06-JUN-76  1434	FTP:Jack Holloway  (H @ MIT-AI)    
Date: 6 JUN 1976 1732-EST
From: Jack Holloway  (H @ MIT-AI)
To: jbr at SU-AI

Hi, we are just trying to bring up a new set of DRAW sources.
What changes have you guys made since January 1, when we got copies
of your NEW,DRW?
-------

∂06-JUN-76  2302	BPM   via AMET	BBN-TENEX (short BBNE)   
Yes, it is up on the ARPAnet.  Host number is 5 (octal or decimal!).
Please tell TELNET, FTP, MAIL, etc. about it.