NETWORKING IN GENERAL
                            OR
           ADDRESSING THE INTEROPERABILITY OF LMI PRODUCTS
        WITH EACH OTHER AND WITH OTHER MANUFACTURERS PRODUCTS.



 BECAUSE IT IS A TENENT OF THE LMI PRODUCT LINE THAT IT
 SUPPORT CLOSE COUPLING OF LISP AND CONVENTIONAL PROCESSORS
 THE AUTHORS PROPOSE THAT LMI CANNOT BASE ITS NETWORKING
 (OR INTEROPERABILITY) STRATEGY ON ANY EXISTING NETWORK
 STANDARD.

 THE LOWEST LEVEL, OR PHYSICAL LINK LEVEL IF YOU WILL, THAT
 LMI SUPPORTS FOLLOWS A MEMORY MODEL OF DATA ORGANIZATION,
 *NOT* THE SERIAL BIT MODEL UPON WHICH EXISTING NETWORK
 STANDARDS ARE CONCEPTUALIZED.

 THE ADVANTAGES OF THE MEMORY MODEL OVER THE BIT SERIAL MODEL
 ARE MANY. [REFER TO RECENT WORK AT CMU AND OTHER PLACES IF
 YOU ACADEMIC DISCUSSIONS OF SUCH THINGS ARE OF YOUR INTEREST.]

 SUMMARY PROPOSAL:
  - TO RECOGNIZE THIS FACT OF THE LMI PRODUCT LINE.
  - TO RECONGIZE THE ADVANTAGEOUS POSITION THIS PUTS LMI IN.
  - TO TAKE ADVANTAGE OF IT BY REORGANIZING SOME OF OUR SYSTEM SOFTWARE
    AROUND IT.

 SOME IMPLICATIONS TO BE SHOWN:

  - LMI SHOULD COME OUT WITH A NUBUS<->NUBUS COUPLER, POSSIBLY
    A FIBER OPTIC TOKEN RING. POSSIBLY THE OLD BUS-COUPLER THAT
    WAS DEMONSTRATABLY WORKING A COUPLE YEARS AGO COULD BE
    MANUFACTURED TO FILL ANY MARKETTING GAPS IN COMING OUT WITH
    A BRAND NEW ADVANCED-TECHNOLOGY COUPLER.

  - MOBY ADDRESS SCHEME CAN BE GENERALIZED TO ENCOMPASS DATA
    FROM NON-LISP SYSTEM ENVIRONMENTS SO LONG AS THOSE ENVIRONMENTS
    STILL ADHERE TO AN OBJECT-ORIENTED VIRTUAL-MEMORY DISCIPLINE.
    [EXAMPLES: IBM SYSTEM-38, SOME ADA ENVIRONMENTS]

  - SHOULD BE ABLE TO BUILD A LISP BAND WITHOUT ANY CONVENTIONAL
    (SERIAL ORIENTED) SOFTWARE.

  - COMMUNICATIONS WITH IO/PROCESSOR, UNIX, MS-DOS, SHOULD
    USE A MEMORY MODEL WITH COMMON DATA FORMATS FOR BASIC MESSAGES
    AND PROTOCOLS.

  - ALL AREAS OF I/O SHOULD BE COVERED.
    * SCREEN/GRAPHICS
    * DISK
    * NETWORK
    * INTERPROCESSOR