Network Working Group                                       Bob Anderson
Request for Comments: 166                                           Rand
NIC 6780                                                       Vint Cerf
                                                                    UCLA
                                                            Eric Harslem
                                                            John Haefner
                                                                    Rand
                                                              Jim Madden
                                                          U. of Illinois
                                                            Bob Metcalfe
                                                                     MIT
                                                           Arie Shoshani
                                                                     SDC
                                                               Jim White
                                                                    UCSB
                                                              David Wood
                                                                   Mitre
                                                             25 May 1971

    DATA RECONFIGURATION SERVICE -- AN IMPLEMENTATION SPECIFICATION

                                 CONTENTS

     I.  INTRODUCTION ...................................  2

         Purpose of this RFC ............................  2
         Motivation .....................................  2

    II.  OVERVIEW OF THE DATA RECONFIGURATION SERVICE ...  3

         Elements of the Data Reconfiguration SERVICE ...  3
         Conceptual Network Connections .................  3
         Conception Protocols and Message Formats .......  4
         Example Connection Configurations ..............  7

   III.  THE FORM MACHINE ...............................  8

         Input/Output Streams and Forms .................  8
         Form Machine BNF Syntax ........................  8
         Alternate Specification of Form Machine Syntax .  9
         Forms .......................................... 10
         Rules .......................................... 10
         Terms .......................................... 11

           Term Format 1 ................................ 11
           Term Format 2 ................................ 11
           Term Format 3 ................................ 14
           Term Format 4 ................................ 14



Anderson, et al.                                                [Page 1]


RFC 166               Data Reconfiguration Service              May 1971


           The Application of a Term .................... 14
           Restrictions and Interpretations of Term
             Functions .................................. 15

           Term and Rule Sequencing ..................... 16

    IV.  EXAMPLES ....................................... 17

         Remarks ........................................ 17
         Field Insertion ................................ 17
         Deletion ....................................... 17
         Variable Length Records ........................ 18
         String Length Computation ...................... 18
         Transposition .................................. 18
         Character Packing and Unpacking ................ 18


                             I.  INTRODUCTION

PURPOSE OF THIS RFC

   The Purpose of this RFC is to specify the Data Reconfiguration
   Service (DRS.)

   The DRS experiment involves a software mechanism to reformat Network
   data streams.  The mechanism can be adapted to numerous Network
   application programs.  We hope that the result of the experiment will
   lead to a future standard service that embodies the principles
   described in this RFC.


MOTIVATION

   Application programs require specific data I/O formats yet the
   formats are different from program to program.  We take the position
   that the Network should adapt to the individual program requirements
   rather than changing each program to comply with a standard.  This
   position doesn't preclude the use of standards that describe the
   formats of regular message contents; it is merely an interpretation
   of a standard as being a desirable mode of operation but not a
   necessary one.

   In addition to differing program requirements, a format mismatch
   problem occurs where users wish to employ many different kinds of
   consoles to attach to a single service program.  It is desirable to
   have the Network adapt to individual console configurations rather
   than requiring unique software packages for each console
   transformation.



Anderson, et al.                                                [Page 2]


RFC 166               Data Reconfiguration Service              May 1971


   One approach to providing adaptation is for those sites with
   substantial computing power to offer a data reconfiguration service;
   this document is a specification of such a service.

   The envisioned modus operandi of the service is that an applications
   programmer defines _forms_ that describe data reconfigurations.  The
   service stores the forms by name.  At a later time, a user (perhaps a
   non-programmer) employs the service to accomplish a particular
   transformation of a Network data stream, simply by calling the form
   by name.

   We have attempted to provide a notation tailored to some specifically
   needed instances of data reformatting while keeping the notation and
   its underlying implementation within some utility range that is
   bounded on the lower end by a notation expressive enough to make the
   experimental service useful, and that is bounded on the upper end by
   a notation short of a general purpose programming language.


             II.  OVERVIEW OF THE DATA RECONFIGURATION SERVICE

ELEMENTS OF THE DATA RECONFIGURATION SERVICE

   An implementation of the Data Reconfiguration Service (DRS) includes
   modules for connection protocols, a handler of some requests that can
   be made of the service, a compiler and/or interpreter (called the
   Form Machine) to act on those requests, and a file storage module for
   saving and retrieving definitions of data reconfigurations (forms).

   This section describes connection protocols and requests.  The next
   section covers the Form Machine language in some detail.  File
   storage is not described in this document because it is transparent
   to the use of the service an its implementation is different at each
   DRS host.

CONCEPTUAL NETWORK CONNECTIONS

   There are three conceptual Network connections to the DRS, see Fig.
   1.

         1)  The control connection (CC) is between an originating user
             and the DRS.  Forms specifying data reconfigurations are
             defined over this connection.  The user indicates (once)
             forms to be applied to data passing over the two
             connections described below.

         2)  The user connection (UC) is between a user process and the
             DRS.



Anderson, et al.                                                [Page 3]


RFC 166               Data Reconfiguration Service              May 1971


         3)  The server connection (SC) is between the DRS and the
             serving process.

   Since the goal is to adapt the Network to user and server processes,
   a minimum of requirements are imposed on the UC and SC.

      +------------+              +------+          +---------+
      | ORIGINATING|     CC       | DRS  |    SC    | SERVER  |
      | USER       |--------------|      |----------| PROCESS |
      +------------+     ^        +------+     ^    +---------+
                         |           /         |
                         |        UC/ <-----\  |
                         |         /         \ |
                         |   +-----------+    \|
         TELNET ---------+   | USER      |     +-- Simplex or Duplex
         Protocol            | PROCESS   |         Connections
         Connection          +-----------+


                Figure 1.  DRS Network Connections


CONNECTION PROTOCOLS AND MESSAGE FORMATS

   Over a control connection the dialog is directly between an
   originating user and the DRS.  Here the user is defining forms or
   assigning predefined forms to connections for reformatting.

   The user connects to the DRS via the standard initial connection
   protocol (ICP).  Rather than going through a logger, the user calls
   on a particular socket on which the DRS alway listens. (Experimental
   socket numbers will be published later.) DRS switches the user to
   another socket pair.

   Messages sent over a control connection are of the types and formats
   specified for TELNET.  (The data type code should specify ASCII --
   the default.)  Thus, a user at a terminal should be able to connect
   to a DRS via his local TELNET, for example, as shown in Fig. 2.

                            +---------+   CC  +---------+
                  +---------| TELNET  |-------|   DRS   |
                  |         +---------+       +---------+
      +-----------------------+
      |         USER          |
      | (TERMINAL OR PROGRAM) |
      +-----------------------+

                  Figure 2. A TELNET Connection to DRS



Anderson, et al.                                                [Page 4]


RFC 166               Data Reconfiguration Service              May 1971


   When a user connects to DRS he supplies a six-character user ID (UID)
   as a qualifier to guarantee the uniqueness of his form names.  He
   will initially have the following commands:

         1.  DEFFORM (form)
         2.  ENDFORM (form)

             These two commands define a form, the text of which is
             chronologically entered between them.  The form is stored
             in the DRS local file system.

         3.  PURGE (form)

             The named form, as qualified by the current UID, is purged
             from the DRS file system.

         4.  LISTNAMES (UID)

             The unqualified names of all forms assigned to UID are
             returned.

         5.  LISTFORM (form)

             The source text of a named form is returned.

         6.  DUPLEXCONNECT (user site, user receive socket, user method,
             server site, server receive socket, server method, user-
             to-server form name, server-to-user form name)

             A duplex connection is made between two processes using the
             receive sockets and the sockets one greater.  Method is
             defined below.  The forms define the transformations on
             these connections.

         7.  SIMPLEXCONNECT (user site, user socket, user method, server
             site, server socket, server method, form)

             A simplex connection is made between the two sockets as
             specified by method.

         8.  ABORT (site, receive socket)

             The reconfiguration of data is terminated by closing both
             the UC and SC specified in part in the command.

   Either one, both, or neither of the two parties specified in 6 or 7
   may be at the same host as the party issuing the request.  Sites and
   sockets specify user and server for the connection.  Method indicates



Anderson, et al.                                                [Page 5]


RFC 166               Data Reconfiguration Service              May 1971


   the way in which the connection is established.

   The following rules apply to these commands:

         1)  Commands may be abbreviated to the minimum number of
             characters to identify them uniquely.

         2)  All commands should be at the start of a line.

         3)  Parameters are enclosed in parentheses and separated by
             commas.

         4)  Imbedded blanks are ignored.

         5)  The parameters are:

             form name        1-6 characters
             UID              1-6 characters
             Site             1-2 characters specifying
                                  the hexadecimal host number
             Socket           1-8 characters specifying the
                                  hexadecimal socket number
             Method           A single character

         6)  Method has the following values:

             C      The site/socket is already connected
                    to the DRS as a dummy control connection
                    (should not be the real control connection).
             I      Connect via the standard ICP (does not
                    apply to SIMPLEXCONNECT).
             D      Connect directly via STR, RTS.

             The DRS will make at least the following minimal
             responses to the user:

             1)  A positive or negative acknowledgement after
                 each line (CR/LF)
             2)  If a form fails or terminates
             TERMINATE, ASCII Host # as hex, ASCII Socket # as hex,
                         ASCII Return Code as decimal
             thus identifying at least one end of the connection.









Anderson, et al.                                                [Page 6]


RFC 166               Data Reconfiguration Service              May 1971


EXAMPLE CONNECTION CONFIGURATIONS

   There are basically two modes of DRS operation: 1) the user wishes to
   establish a DRS UC/SC connection(s) between the programs and 2) the
   user wants to establish the same connection(s) where he (his
   terminal) is at the end of the UC or the SC.  The latter case is
   appropriate when the user wishes to interact from his terminal with
   the serving process (e.g., a logger).

   In the first case (Fig. 1, where the originating user is either a
   terminal or a program) the user issues the appropriate CONNECT
   command.  The UC/SC can be simplex or duplex.

   The second case has two possible configurations, shown in Figs. 3 and
   4.


   +-------+    +--------+   CC    +-----+        +----+
   |       |----|        |---------|     |   SC   |    |
   | USER  |    | TELNET |   UC    | DRS |--------| SP |
   |       |----|        |---------|     |        |    |
   +-------+    +--------+         +-----+        +----+

            Figure 3.  Use of Dummy Control Connection

                +---------+
   +------+    /| USER    |   CC   +-----+
   |      |---/ | SIDE    |--------|     |   SC   +----+
   | USER |     +---------+   UC   | DRS |--------| SP |
   |      |---\ | SERVING |--------|     |        +----+
   +------+    \| SIDE    |        +-----+
                +---------+

            Figure 4.  Use of Server TELNET

   In Fig. 3 the user instructs his TELNET to make two duplex
   connections to DRS.  One is used for control information (the CC) and
   the other is a dummy.  When he issues the CONNECT he references the
   dummy duplex connection (UC) using the "already connected" option.

   In Fig. 4 the user has his TELNET (user side) call the DRS.  When he
   issues the CONNECT the DRS calls the TELNET (server side) which
   accepts the call on behalf of the console.  This distinction is known
   only to the user since to the DRS the configuration Fig. 4 appears
   identical to that in Fig. 1.  Two points should be noted:

        1)  TELNET protocol is needed only to define forms and direct
            connections.  It is not required for the using and serving



Anderson, et al.                                                [Page 7]


RFC 166               Data Reconfiguration Service              May 1971


            processes.
        2)  The using and serving processes need only a minimum of
            modification for Network use, i.e., an NCP interface.

                          III.  THE FORM MACHINE

INPUT/OUTPUT STREAMS AND FORMS

   This section describes the syntax and semantics of forms that specify
   the data reconfigurations.  The Form Machine gets an input stream,
   reformats the input stream according to a form describing the
   reconfiguration, and emits the reformatted data as an output stream.

   In reading this section it will be helpful to envision the
   application of a form to the data stream as depicted in Fig. 5.  An
   input stream pointer identifies the position of data (in the input
   stream) that is being analyzed at any given time by a part of the
   form.  Likewise, an output stream pointer locates data being emitted
   in the output stream.


       /\/\                                                  /\/\
  ^    |  |                     FORM                         |  |   ^
  |    |  |                -----------------                 |  |   |
  |    |  |            +-  -----------------  -+             |  |   |
  |    |  |            |   CURRENT PART OF     |             |  |   |
INPUT  |  |<= CURRENT <    -----------------    > CURRENT => |  | OUTPUT
STREAM |  |   POINTER  |   FORM BEING APPLIED  |  POINTER    |  | STREAM
       |  |            +-  -----------------  -+             |  |
       |  |                -----------------                 |  |
       |  |                -----------------                 |  |
       |  |                -----------------                 |  |
       \/\/                                                  \/\/
              Figure 5.  Application of Form to Data Streams

















Anderson, et al.                                                [Page 8]


RFC 166               Data Reconfiguration Service              May 1971


FORM MACHINE BNF SYNTAX

   form           ::=  rule | rule form

   rule           ;;=  label  inputstream  outputstream ;

   label          ::=  INTEGER | <null>

   inputstream    ::=  terms | <null>

   terms          ::=  term | terms , term

   outputstream   ::=  : terms | <null>

   term           ::=  identifier | identifier  descriptor |
                       descriptor | comparator

   identifier     ::=  an alpha character followed by 0 to 3
                       alphanumerics

   descriptor     ::=  (replicationexpression , datatype ,
                       valueexpression , lengthexpression  control)

   comparator     ::=  (value  connective  value  control)  |
                       (identifier  *<=*  control)

   replicationexpression  ::=  # | arithmeticexpression | <null>

   datatype       ::=  B | O | X | E | A

   valueexpression  ::=  value | <null>

   lengthexpression  ::=      arithmeticexpression | <null>

   connective     ::=  .LE. | .LT. | .GE. | .GT. | .EQ. | .NE.

   value          ::=  literal | arithmeticexpression

   arithmeticexpression  ::=  primary | primary operator
                              arithmeticexpression

   primary        ::=  identifier | L(identifier) | V(identifier) |
                       INTEGER

   operator       ::=  + | - | * | /

   literal        ::=  literaltype "string"




Anderson, et al.                                                [Page 9]


RFC 166               Data Reconfiguration Service              May 1971


   literaltype    ::=  B | O | X | E | A

   string         ::=  from 0 to 256 characters

   control        ::=  :  options | <null>

   options        ::=  S(where) | F(where) | U(where) |
                       S(where) , F(where) |
                       F(where) , S(where)

   where          ::=  arithmeticexpression | R(arithmeticexpression)


ALTERNATE SPECIFICATION OF FORM MACHINE SYNTAX

                                   infinity
form                    ::=  {rule}
                                   1
                                      1         1          1
rule                    ::=  {INTEGER}   {terms}   {:terms} ;
                                      0         0          0
                                         infinity
terms                   ::=  term {,term}
                                         0
                                                      1
term                    ::=  identifier | {identifier}   descriptor
                                                      0
                             | comparator
                                                    1
descriptor              ::=  ({arithmeticexpression}  , datatype ,
                                                    0
                                    1                     1          1
                             {value} ,  {lengthexpression}  {:options}
                                    0                     0          0
                                                                 1
comparator              ::=  (value  connective  value {:options} ) |
                                                                 0
                                                              1
                             (identifier .<=. value {:options} )
                                                              0
connective              ::=  .LE. | .LT. | .GE. | .GT. | .EQ. | .NE.

lengthexpression        ::=  # | arithmeticexpression

datatype                ::=  B | O | X | E | A

value                   ::=  literal | arithmeticexpression




Anderson, et al.                                               [Page 10]


RFC 166               Data Reconfiguration Service              May 1971


                                                         infinity
arithmeticexpression    ::=  primary  {operator  primary}
                                                         0
operator                ::= + | - | * | /

primary                 ::=  identifier | L(identifier) |
                             V(identifier) | INTEGER
                                                      256
literal                 ::=  literaltype  "{CHARACTER}   "
                                                      0
literaltype             ::=  B | O | X | A | E
                                                 1
options                 ::=  S(where) {,F(where)}  |
                                                 0
                                                 1
                             F(where) {,S(where)}  | U(where)
                                                 0
where                   ::=  arithmeticexpression |
                             R(arithmeticexpression)
                                                     3
identifier              ::=  ALPHABETIC  {ALPHAMERIC}
                                                     0


FORMS

   A form is an ordered set of rules.

         form ::=  rule | rule form

   The current rule is applied to the current position of the input
   stream.  If the (input stream part of a) rule fails to correctly
   describe the contents of the current input then another rule is made
   current and applied to the current position of the input stream.  The
   next rule to be made current is either explicitly specified by the
   current term in the current rule or it is the next sequential rule by
   default.  Flow of control is more fully described under TERM AND RULE
   SEQUENCING.

   If the (input stream part of a) rule succeeds in correctly describing
   the current input stream, then some data may be emitted at the
   current position in the output stream according to the rule.  The
   input and output stream pointers are advanced over the described and
   emitted data, respectively, and the next rule is applied to the now
   current position of the input stream.

   Application of the form is terminated when an explicit return
   (R(arithmeticexpression)) is encountered in a rule.  The user and



Anderson, et al.                                               [Page 11]


RFC 166               Data Reconfiguration Service              May 1971


   server connections are closed and the return code
   (arithmeticexpression) is sent to the originating user.

RULES

   A rule is a replacement, comparison, and/or an assignment operation
   of the form shown below.

         rule ::= label  inputstream  outputstream

   A label is the name of a rule and it exists so that the rule may be
   referenced elsewhere in the form for explicit rule transfer of
   control.  Labels are of the form below.

         label ::=  INTEGER | <null>

   The optional integer labels are in the range 0 >= INTEGER >= 9999.
   The rules need not be labeled in ascending numerical order.

TERMS

   The inputstream (describing the input stream to be matched) and the
   outputstream (describing data to be emitted in the output stream)
   consist of zero or more terms and are of the form shown below.

         inputstream   ::=  terms | <null>
         outputstream  ::=  :terms | <null>
         terms         ::=  term | terms , term

   Terms are of one of four formats as indicated below.

         term ::=  identifier | identifier  descriptor |
                   descriptor | comparator

Term Format 1

   The first term format is shown below.

         identifier

   The identifier is a symbolic reference to a previously identified
   term (term format 2) in the form.  It takes on the same attributes
   (value, length, type) as the term by that name.  Term format 1 is
   normally used to emit data in the output stream.

   Identifiers are formed by an alpha character followed by 0 to 3
   alphanumeric characters.




Anderson, et al.                                               [Page 12]


RFC 166               Data Reconfiguration Service              May 1971


Term Format 2

   The second term format is shown below.

         identifier descriptor

   Term format 2 is generally used as an input stream term but can be
   used as an output stream term.

   A descriptor is defined as shown below.

         descriptor ::= (replicationexpression, datatype,
                        valueexpression, lengthexpression
                        control)

   The identifier is the symbolic name of the term in the usual
   programming language sense.  It takes on the type, length, value, and
   replication attributes of the term and it may be referenced elsewhere
   in the form.

   The replication expression, if specified, causes the unit value of
   the term to be generated the number of times indicated by the value
   of the replication expression.  The unit value of the term (quantity
   to be replicated) is determined from the data type, value expression,
   and length expression attributes.  The data type defines the kind of
   data being specified.  The value expression specifies a nominal value
   that is augmented by the other term attributes.  The length
   expression determines the unit length of the term.  (See the IBM SRL
   Form C28-6514 for a similar interpretation of the pseudo instruction,
   defined constant, after which the descriptor was modeled.)

   The replication expression is defined below.

         replicationexpression ::= # | arithmeticexpression | <null>
         arithmeticexpression ::= primary | primary operator
                                  arithmeticexpression
         operator ::= + | - | * | /
         primary ::= identifier | L(identifier) | V(identifier) |
                     INTEGER

   The replication expression is a repeat function applied to the
   combined data type value, and length expressions.  It expresses the
   number of times that the nominal value is to be repeated.

   The terminal symbol # means an arbitrary replication factor.  It must
   be explicitly terminated by a match or non-match to the input stream.
   This termination may result from the same or the following term.




Anderson, et al.                                               [Page 13]


RFC 166               Data Reconfiguration Service              May 1971


   A null replication expression has the value of one.  Arithmetic
   expressions are evaluated from left-to-right with no precedence.

   The L(identifier) is a length operator that generates a 32-bit binary
   integer corresponding to the length of the term named.  The
   V(identifier) is a value operator that generates a 32-bit binary
   integer corresponding to the value of the term named.  (See
   Restrictions and Interpretations of Term Functions.)  The value
   operator is intended to convert character strings to their numerical
   correspondents.

   The data type is defined below.

             datatype ::= B | O | X | E | A

   The data type describes the kind of data that the term represents.
   (It is expected that additional data types, such as floating point
   and user-defined types, will be added as needed.)

        Data Type         Meaning              Unit Length

            B             Bit string              1 bit
            O             Bit string              3 bits
            X             Bit string              4 bits
            E             EBCDIC character        8 bits
            A             Network ASCII character 8 bits

   The value expression is defined below.

            valueexpression ::= value | <null>
            value ::= literal | arithmeticexpression
            literal ::= literaltype "string"
            literaltype ::= B | O | X | E | A

   The value expression is the nominal value of a term expressed in the
   format indicated by the data type.  It is repeated according to the
   replication expression.

   A null value expression in the input stream defaults to the data
   present in the input stream.  The data must comply with the datatype
   attribute, however.

   A null value expression generates padding according to Restrictions
   and Interpretations of Term Functions.

   The length expression is defined below.

         lengthexpression ::= arithmeticexpression | <null>



Anderson, et al.                                               [Page 14]


RFC 166               Data Reconfiguration Service              May 1971


   The length expression states the length of the field containing the
   value expression.

   If the length expression is less than or equal to zero, the term
   succeeds but the appropriate stream pointer is not advanced.
   Positive lengths cause the appropriate stream pointer to be advanced
   if the term otherwise succeeds.

   Control is defined under TERM AND RULE SEQUENCING.

Term Format 3

   Term format 3 is shown below.

         descriptor

   It is identical to term format 2 with the omission of the identifier.
   Term format 3 is generally used in the output stream.  It is used in
   the input stream where input data is to be passed over but not
   retained for emission or later reference.

Term Format 4

   The fourth term format is shown below.

         comparator    ::= (value connective value control) |
                           (identifier *<=* value  control)
         value         ::= literal | arithmeticexpression
         literal       ::= literaltype "string"
         literaltype   ::= B | O | X | E | A
         string        ::= from 0 to 256 characters
         connective    ::= .LE. | .LT. | .GE. | .GT. | .EQ. | .NE.

   The fourth term format is used for assignment and comparison.

   The assignment operator *<=* assigns the value to the identifier.
   The connectives have their usual meaning.  Values to be compared must
   have the same type and length attributes or an error condition arises
   and the form fails.

The Application of a Term

   The elements of a term are applied by the following sequence of
   steps.

         1.  The data type, value expression, and length expression
             together specify a unit value, call it x.




Anderson, et al.                                               [Page 15]


RFC 166               Data Reconfiguration Service              May 1971


         2.  The replication expression specifies the number of times x
             is to be repeated.  The value of the concatenated xs
             becomes y of length L.

         3.  If the term is an input stream term then the value of y of
             length L is tested with the input value beginning at the
             current input pointer position.

         4.  If the input value satisfies the constraints of y over
             length L then the input value of length L becomes the value
             of the term.

   In an output stream term, the procedure is the same except that the
   source of input is the value of the term(s) named in the value
   expression and the data is emitted in the output stream.

   The above procedure is modified to include a one term look-ahead
   where replicated values are of indefinite length because of the
   arbitrary symbol, #.

Restrictions and Interpretations of Term Functions

   1.    Terms having indefinite lengths because their values are
         repeated according to the # symbol, must be separated by some
         type-specific data such as a literal.  (A literal isn't
         specifically required, however.  An arbitrary number of ASCII
         characters could be terminated by a non-ASCII character.)

   2.    Truncation and padding is as follows:
         a)  Character to character (A <-> E) conversion is left-
             justified and truncated or padded on the right with blanks.
         b)  Character to numeric and numeric to numeric conversions are
             right-justified and truncated or padded on the left with
             zeros.
         c)  Numeric to character conversions is right-justified and
             left-padded with blanks.

   3.    The following are ignored in a form definition over the control
         connection.
         a)  TELNET control characters.
         b)  Blanks except within quotes.
         c)  /* string */ is treated as comments except within quotes.

   4.    The following defaults prevail where the term part is omitted.

         a)  The replication expression defaults to one.
         b)  # in an output stream term defaults to one.
         c)  The value expression of an input stream term defaults to



Anderson, et al.                                               [Page 16]


RFC 166               Data Reconfiguration Service              May 1971


             the value found in the input stream, but the input stream
             must conform to the data type and length expression.  The
             value expression of an output stream term defaults to
             padding only.
         e)  The length expression defaults to the size of the quantity
             determined by the data type and value expression.
         f)  Control defaults to the next sequential term if a term is
             successfully applied; else control defaults to the next
             sequential rule.  If _where_ evaluates to an undefined
             _label_ the form fails.

   5.    Arithmetic expressions are evaluated left-to-right with no
         precedence.

   6.    The following limits prevail.

         a)  Binary lengths are <= 32 bits
         b)  Character strings are <= 256 8-bit characters
         c)  Identifier names are <= 4 characters
         d)  Maximum number of identifiers is <= 256
         e)  Label integers are >= 0 and <= 9999
   7.    Value and length operators product 32-bit binary integers.  The
         value operator is currently intended for converting A or E type
         decimal character strings to their binary correspondents.  For
         example, the value of E'12' would be 0......01100.  The value
         of E'AB' would cause the form to fail.

TERM AND RULE SEQUENCING

   Sequencing may be explicitly controlled by including control in a
   term.

        control ::=  :options | <null>
        options ::=  S(where) | F(where) | U(where)
                     S(where) , F(where) |
                     F(where) , S(where)

        where   ::=  arithmeticexpression | R(arithmeticexpression)

   S, F, and U denote success, fail, and unconditional transfers,
   respectively.  _Where_ evaluates to a _rule_ label, thus transfer can
   be effected from within a rule (at the end of a term) to the
   beginning of another rule.  R means terminate the form and return the
   evaluated expression to the initiator over the control connection (if
   still open).

   If terms are not explicitly sequenced, the following defaults
   prevail.



Anderson, et al.                                               [Page 17]


RFC 166               Data Reconfiguration Service              May 1971


        1)  When a term fails go to the next sequential rule.
        2)  When a term succeeds go to the next sequential
            term within the rule.
        3)  At the end of a rule, go to the next sequential
            rule.

   Note in the following example, the correlation between transfer of
   control and movement of the input pointer.

        1   XYZ(,B,,8:S(2),F(3)) : XYZ ;
        2   . . . . . . .
        3   . . . . . . .

   The value of XYZ will never be emitted in the output stream since
   control is transferred out of the rule upon either success or
   failure.  If the term succeeds, the 8 bits of input will be assigned
   as the value of XYZ and rule 2 will then be applied to the same input
   stream data.  That is, since the complete left hand side of rule 1
   was not successfully applied, the input stream pointer is not
   advanced.

                               IV.  EXAMPLES

REMARKS

   The following examples (forms and also single rules) are simple
   representative uses of the Form Machine.  The examples are expressed
   in a term-per-line format only to aid the explanation.  Typically, a
   single rule might be written as a single line.

FIELD INSERTION

   To insert a field, separate the input into the two terms to allow the
   inserted field between them.  For example, to do line numbering for a
   121 character/line printer with a leading carriage control character,
   use the following form.

   (NUMB*<=*1);       /*initialize line number counter to one*/
   1 CC(,E,,1:F(R(99))),  /*pick up control character and save
                          as CC*/
                          /*return a code of 99 upon exhaustion*/
   LINE(,E,,121 : F(R(98)))  /*save text as LINE*/
   :CC,               /*emit control character*/
   (,E,NUMB,2),       /*emit counter in first two columns*/
   (,E,E".",1),       /*emit period after line number*/
   (,E,LINE,117),     /*emit text, truncated in 117 byte field*/
   (NUMB*<=*NUMB+1:U(1));   /*increment line counter and go to
                              rule one*/;;



Anderson, et al.                                               [Page 18]


RFC 166               Data Reconfiguration Service              May 1971


DELETION

   Data to be deleted should be isolated as separate terms on the left,
   so they may be omitted (by not emitting them) on the right.

   (,B,,8),           /*isolate 8 bits to ignore*/
   SAVE(,A,,10)       /*extract 10 ASCII characters from
                        input stream*/
   :(,E,SAVE,);       /*emit the characters in SAVE as EBCDIC
                        characters whose length defaults to the
                        length of SAVE, i.e., 10, and advance to
                        the next rule*/

   In the above example, if either input stream term fails,
   the next sequential rule is applied.

VARIABLE LENGTH RECORDS

   Some devices, terminals and programs generate variable
   length records.  The following rule picks up variable length
   EBCDIC records and translates them to ASCII.

   CHAR(#,E,,1),      /*pick up all (an arbitrary number of)
                        EBCDIC characters in the input stream*/
   (,X,X"FF",2)       /*followed by a hexadecimal literal,
                        FF (terminal signal)*/
   :(,A,CHAR,),       /*emit them as ASCII*/
   (,X,X"25",2);      /*emit an ASCII carriage return*/

STRING LENGTH COMPUTATION

   It is often necessary to prefix a length field to an arbitrarily long
   character string.  The following rule prefixes an EBCDIC string with
   a one-byte length field.

   Q(#,E,,1),         /*pick up all EBCDIC characters*/
   TS(,X,X"FF",2)     /*followed by a hexadecimal literal, FF*/
   :(,B,L(Q)+2,8),    /*emit the length of the characters
                        plus the length of the literal plus
                        the length of the count field itself,
                        in an 8-bit field*/
   Q,                 /*emit the characters*/
   TS,                /*emit the terminal*/








Anderson, et al.                                               [Page 19]


RFC 166               Data Reconfiguration Service              May 1971


TRANSPOSITION

   It is often desirable to reorder fields, such as the following
   example.

   Q(,E,,20), R(,E,,10) , S(,E,,15), T(,E,,5) : R, T, S, Q ;

   The terms are emitted in a different order.

CHARACTER PACKING AND UNPACKING

   In systems such as HASP, repeated sequences of characters are packed
   into a count followed by the character, for more efficient storage
   and transmission.  The first form packs multiple characters and the
   second unpacks them.

   /*form to pack EBCDIC streams*/
   /*returns 99 if OK, input exhausted*/
   /*returns 98 if illegal EBCDIC*/
   /*look for terminal signal FF which is not a legal EBCDIC*/
   /*duplication count must be 0-254*/
   1 (,X,X"FF",2 : S(R(99))) ;
   /*pick up an EBCDIC char/*
   CHAR(,E,,1) ;
   /*get identical EBCDIC chars/*
   LEN(#,E,CHAR,1)
   /*emit the count and the char/*
   : (,B,L(LEN)+1,8), CHAR, (:U(1));
   /*end of form*/;;

   /*form to unpack EBCDIC streams*/
   /*look for terminal*/
   1 (,X,X"FF",2 : S(R(99))) ;
   /*emit character the number of times indicated*/
   /*by the count, in a field the length indicated*/
   /*by the counter contents*/
   CNT(,B,,8), CHAR(,E,,1) : (CNT,E,CHAR,1:U(1));
   /*failure of form*/
   (:U(R(98))) ;;




       [ This RFC was put into machine readable form for entry ]
        [ into the online RFC archives by Simone Demmel 03/98 ]






Anderson, et al.                                               [Page 20]