00100	                         STANFORD UNIVERSITY
00200	                     COMPUTER SCIENCE DEPARTMENT
00300	
00400	
00500	                          January 30, 1973
00600	
00700	
00800	
00900	Dr. Larry Roberts
01000	Advanced Research Projects Agency
01100	1400 Wilson Boulevard
01200	Arlington, Virginia  22202
01300	
01400	Dear Larry:
01500	
01600		Enclosed are a letter and a report from III that I am sending
01700	you  with  their permission.  I think that reading them will convince
01800	you of the following propositions:
01900	
02000		1. Character  reading  of  non-OCR  fonts  is  still  in  the
02100	research stage.
02200	
02300		2. III has made reasonable progress.
02400	
02500		3.  They  are  unlikely to have a commercial success, because
02600	the market isn't there yet.
02700	
02800		In my opinion, undertaking their OCR venture  was  a  mistake
02900	for  a  company,  because  when the market does materialize, they are
03000	unlikely to get the full benefit of their pioneering.  Their  machine
03100	will  be  obsolete,  and  the  techniques  will  not  be that hard to
03200	rediscover.  There will, however, be considerable  loss  of  time  in
03300	rediscovering them.
03400	
03500		There  probably  will be a large market for general character
03600	reading once there are large numbers of display  terminals  and  hard
03700	copy  terminals  and files all connected by a suitable network.  Then
03800	it will pay to keep documents on file, and it will begin  to  pay  to
03900	read  in old documents.  This time is not far off, but it is not here
04000	yet.
04100	
04200		I believe that the Defense  Department  is  one  organization
04300	that  will  have  a  considerable interest in massive text files, and
04400	that they are a necessary part of the ARPA network.
04500	
04600		In view of this, there are several alternatives:
04700	
04800		1. The simplest is to do nothing now.  In this case, the  III
04900	capability  may well be lost, but the work can certainly be done over
05000	when the time is ripe.
05100	
05200		2. What III would undoubtedly prefer is for  ARPA  to  buy  a
05300	GRAPHICS  I  system.  It could then be given to a suitable contractor
05400	as a network facility and for  developing  further  techniques.  This
05500	might  be  fine  if  there  turns  out  to  be  a  suitable candidate
05600	contractor with the right  interests  and  capability,  but  probably
05700	there isn't one.
05800	
05900		3.  The  third alternative is to give III a research contract
06000	large enough to keep the group going.  This could  be  combined  with
06100	putting  their  facility on the network and letting it read documents
06200	as a network service.   Network users could either  pay  III  for  it
06300	with  real  money as they pay BBN for paging boxes (in my opinion the
06400	preferable alternative) or funny money could be made  available  that
06500	would be charged against an ARPA contract with III.
06600	
06700		In my opinion, alternative three is best.
06800	
06900					Sincerely yours,
07000	
07100	
07200					John McCarthy