Configuring Broker for Internationalization

Software internationalization is the process of designing products and services so that they can be adapted easily to a variety of different local languages and cultures. Internationalization within EntireX means internationalization of messages: the incoming and outgoing messages are converted to the desired codepage of the platform in use. This document explains in detail how to configure the broker for character conversion. It covers the following topics:

See also Internationalization with EntireX

Configuring ICU Conversion

Start of instruction setTo configure ICU conversion

  1. In the Broker attribute file, set the service-specific attribute CONVERSION. Examples:

  2. Optionally configure a CONVERSION OPTION to tune error behavior to meet your requirements; see OPTION Values for Conversion.

  3. For the Broker attribute, check if ICU conversion is possible, that is, the attribute ICU-CONVERSION is either

    • not defined, its default is YES

    • set to YES

Start of instruction setTo configure locale string defaults (optional)

  • If the broker's locale string defaults do not match your requirements (see Broker's Locale String Defaults), we recommend you assign suitable locale string defaults for your country and region, see the respective attribute in Codepage-specific Attributes for how to customize the broker's locale string defaults.

Start of instruction setTo customize mapping of locale strings (optional)

  • If the built-in locale string mapping mechanism does not match your requirements, you can assign specific codepages to locale strings. See Broker's Built-in Locale String Mapping and locale-string for information on customizing the mapping of locale strings to codepages.

Building and Installing ICU Custom Converters

User-written ICU custom-converters can be used for ACI-based Programming, RPC-based Components, and Reliable RPC. This section covers the following topics:

Writing a User-written ICU Converter

ICU uses algorithmic conversion, non-algorithmic conversion and combinations of both. See ICU Conversion. Non-algorithmic converters defined by the UCM format are the easiest way to define user-written ICU converters. See UCM Format.

Start of instruction setTo write a (non-algorithmic) user-written ICU converter

  • Define the ICU converter file in UCM format using a text editor to meet your requirements.

    Note:
    For further explanation of the UCM file format, see ICU Resources.

Writing algorithmic and partially algorithmic converters can be complex. However, they can be installed into EntireX in the same way as the table-driven, non-algorithmic ones. A description of how to write algorithmic and partially algorithmic converters is beyond the scope of this documentation; please see the ICU documentation and other sources specified under ICU Resources.

Compiling a User-written ICU Converter

Start of instruction setTo compile the user-written ICU converter

  • Compile the converter source files (extension .ucm) into binary converter files (extension ".cnv") using the ICU tool makeconv. Example: 

    makeconv -v myebcdic.ucm

    Note:
    EntireX delivers the ICU tool makeconv in the EntireX bin directory.

    This produces a binary converter file named myebcdic.cnv.

    Caution:
    The binary format "cnv" depends on the endianness (big/little-endian) and character set family (ASCII/EBCDIC) of the computer where it is produced. For example, a binary converter file produced on a machine with big endianes cannot be executed on a machine with little-endian (and vice versa) or character set family EBCDIC cannot be executed on a machine with character set family ASCII (and vice versa). It is highly recommended to compile the converter source file(s) on the same target platform where the broker runs - otherwise unpredictable result may occur.

Installing a User-written ICU Converter

Start of instruction setTo install the user-written ICU converter

  1. Define the broker attribute ICU-DATA-DIRECTORY. See Broker-specific Attributes.

    Example:

    ICU DATA DIRECTORY=".../EntireX/config/etb"

  2. Define the subdirectory icudt<icu-version><endianness> within the ICU-DATA-DIRECTORY

    where <icu-version> is the ICU version used, for example 54, and
      <endianness> is either "b" (big-endian) or "l" (little-endian)

    Examples:

    .../EntireX/config/etb/icudt54l
.../EntireX/config/etb/icudt54b

    Notes:

    1. The subdirectory and its naming are given by ICU standard. It is not invented by Software AG.
    2. See the Release Notes to determine the ICU version used by the broker you are running and form the correct directory name - otherwise the user-written ICU converter will not be located.
    3. Take care to use the correct endianness given by the machine the broker is running on, otherwise the user-written ICU converter will not be located.
    4. There are also other approaches supported by ICU to locate converters. These approaches are (also) ICU version dependent. However, Software AG recommends the mechanism described above. See the ICU website for more information under ICU Resources.

  3. Copy the user-written ICU converter binary file (extension "cnv") to the directory referenced by ICU-DATA-DIRECTORY and its subdirectory defined under steps 1 and 2 above. Examples: 

    .../EntireX/config/etb/icudt54l/myebcdic.cnv
.../EntireX/config/etb/icudt54l/myascii.cnv

  4. If the converter name is not sent as the locale string by your application, customize the mapping of locale strings by assigning the user-written ICU converter (codepage) to locale strings in the Broker attribute file, see locale-string for how to customize the mapping of locale strings to codepages. Example: 

    DEFAULTS=CODEPAGE 
/* Customer-written ICU converter */
CP1140=myebcdic
CP0819=myascii

  5. For the Broker attribute, check whether ICU conversion is possible, that is, the attribute ICU-CONVERSION is not defined (default=YES) or set to YES.

  6. For the Broker attribute, check whether use of ICU custom converters is possible, that is, the attribute ICU-SET-DATA-DIRECTORY is not defined (default=YES) or set to YES.

Writing Translation User Exits

This section covers the following topics:

Introduction

EntireX Broker provides an interface to enable user-written translation routines in the programming language C. It contains three parameters:

  • The address of the TRAP control block (TRAP = Translation Routine / Area for Parameters).

  • The address of a temporary work area. It is aligned to fullword / long integer boundary (divisible by 4). The work area can only be used for temporary needs and is cleared after return.

  • A fullword (long integer) that contains the length of the work area.

Note:
Names for user-written translation routines starting with "SAG" are reserved for Software AG usage and must not be used, e.g. "SAGTCHA" and "SAGTRPC".

Structure of the TRAP Control Block

The C structure TR_TRAP covers the layout of the control block.

typedef struct _TR_TRAP                                     /* I / O */
{
  unsigned long  tr_type;              /* TRAP type: TRAP_TYPE  inp  */
#define TR_TYPE  2                     /* TRAP type ETB 121          */
  long           tr_ilen;              /* Input buffer length   inp  */
  unsigned char *tr_ibuf;              /* Ptr to input buffer   inp  */
  long           tr_olen;              /* Output buffer length  inp  */
  unsigned char *tr_obuf;              /* Ptr to output buffer  inp  */
  long           tr_dlen;              /* Len of data returned: out  */
                                       /* Minimum of tr_ilen         */
                                       /*        and tr_olen         */
  unsigned long  tr_shost;             /* Senders host          inp  */
#define TR_LITTLE_ENDIAN  0            /* little endian              */
#define TR_BIG_ENDIAN     1            /* big endian                 */
  unsigned long  tr_scode;             /* Senders character set inp  */
#define SEBCIBM ((1L << 5)|(1L << 1))  /* 0x22 EBCDIC (IBM)          */
#define SEBCSNI ((1L << 6)|(1L << 1))  /* 0x42 EBCDIC (SNI)          */
#define SA88591            (1L << 7)   /* 0x80 ASCII                 */
  unsigned long  tr_rhost; /* Receivers host     (see tr_shost) inp  */
  unsigned long  tr_rcode; /* Receivers char set (see tr_scode) inp  */
  unsigned long  tr_bhost; /* BROKER host        (see tr_shost) inp  */
  unsigned long  tr_bcode; /* BROKER char set    (see tr_scode) inp  */
  unsigned long  tr_senva; /* Senders ENVIRONMENT field set:    inp  */
#define OFF  0                         /*  ENVIRONMENT field not set */
#define ON   1                         /*  ENVIRONMENT field set     */
  unsigned long  tr_renva; /* Receivers ENVIRONMENT field set:  inp  */
                           /* see tr_senva                           */
#define S_ENV 32           /* size of ENVIRONMENT field              */
  char           tr_senv[S_ENV];/* Senders   ENVIRONMENT field  inp  */
  char           tr_renv[S_ENV];/* Receivers ENVIRONMENT field  inp  */
} TR_TRAP;

Using the TRAP Fields

The tr_dlen must be supplied by the user-written translation routine. It tells the Broker the length of the message of the translation. In our example its value is set to the minimum length of the input and output buffer.

All other TRAP fields are supplied by the Broker and must not be modified by the user-written translation routine.

The incoming message is located in a buffer pointed to by tr_ibuf. The length (not to be exceeded) is supplied in tr_ilen. The character set information from the send buffer can be taken from tr_scode.

The outgoing message must be written to the buffer pointed to by tr_obuf. The length of the output buffer is given in the field tr_olen. The character set is specified in tr_rcode. If the addresses given in tr_ibuf and tr_obuf point to the same location, it is not necessary to copy the data from the input buffer to the output buffer.

The environment fields tr_senva and tr_renva are provided to handle site-dependent character set information. For the SEND and/or RECEIVE functions, you can specify data in the ENVIRONMENT field of the Broker ACI control block. This data is translated into the codepage of the platform where EntireX Broker is running (see field tr_bcode) and is available to the tr_senv or tr_renv field in the TRAP control block. tr_senva or tr_renva are set to ON if environmental data is available. Any values given in the API field ENVIRONMENT must correspond to the values handled in the translation routine.

Configuring Translation User Exits

Start of instruction setTo configure translation user exits

As a prerequisite, the user-written translation routine shared library/object must be accessible to the Broker worker threads.

  1. Copy the user-written translation routine shared library/object into the EntireX lib directory.

  2. In the Broker attribute file, set the service-specific attribute TRANSLATION to the name of the user-written translation routine. Example: 

    TRANSLATION=libmytrans.s[o|l]

    or

  1. Place the user-written translation routine shared library/object in a directory of your choice. Spaces in the path name are not allowed.

  2. In the Broker attribute file, set the service-specific attribute TRANSLATION to the full path name of the directory of the user-written translation routine. Example: 

    TRANSLATION=../mydir/mytrans/libmytrans.s[o|l]

Writing SAGTRPC User Exits

This section covers the following topics:

Introduction

EntireX Broker provides an interface to SAGTRPC user exit routines written in the programming language C. The interface contains three parameters:

  • The address of the UE (user exit) control block.

  • The address of a temporary work area. It is aligned to a fullword / long-integer boundary (divisible by 4). The work area can only be used temporarily and is cleared after return.

  • A fullword (long integer) that contains the length of the work area.

Note:
Names for conversion routines starting with "SAG" are reserved for Software AG usage and must not be used, e.g. "SAGTCHA" and "SAGTRPC".

Structure of the User Exit Control Block

The C structure UECB shows the layout of the user exit control block. 

typedef struct _UECB
{
   unsigned long          eVersion;
#define USRTRPC_VERSION_1                1

   char                *  pInputBuffer;
   unsigned long          uInputLen;
   char                *  pOutputBuffer;
   unsigned long          uOutputLen;
   unsigned long          uReturnedLen;

   unsigned long          shost; 
#define USRTRPC_LITTLE_ENDIAN  0      /* little endian                      */
#define USRTRPC_BIG_ENDIAN     1      /* big endian                         */

   unsigned long          scode;
#define USRTRPC_SEBCIBM ((1L << 5)|(1L << 1)) /* 0x22 EBCDIC (IBM)          */
#define USRTRPC_SEBCSNI ((1L << 6)|(1L << 1)) /* 0x42 EBCDIC (SNI)          */
#define USRTRPC_SA88591            (1L << 7)  /* 0x80 ASCII                 */

   unsigned long          rhost; 
/* see shost */
   unsigned long          rcode;
/* see scode */
   unsigned long          bhost;
/* see shost */
   unsigned long          bcode;
/* see scode */

   unsigned long          uCpSender;
   unsigned long          uCpReceiver;
   unsigned long          uCpBroker;

   char                   eFunction;
#define USRTRPC_FCT_CONVERT             'C'
#define USRTRPC_FCT_GETLENGTH           'L'

   char                   eDirection;
#define USRTRPC_DIR_SENDER_TO_BROKER    '1'
#define USRTRPC_DIR_SENDER_TO_RECEIVER  '2'
#define USRTRPC_DIR_BROKER_TO_RECEIVER  '3'

   char                   sFormat[2];
#define ERX_USERDATA    "01"    /* UserId, Lib, Pgm, etc. from Header 
                                  (truncatable)                         */
#define ERX_METADATA    "02"    /* Header Data (non-truncatable)        */
#define ERX_FRMTDATA    "03"    /* Format Buffer (non-truncatable)      */
#define ERX_SB_ELEMENT  "04"    /* String Buffer                        */
#define ERX_VB_METADATA "05"    /* Value Buffer Array Occurences, 
                                   String Length                        */
#define ERX_PREVIEW     "99"    /* Previewing FB and VB, etc...         */
                                /* Convert data lazy. Do not care on    */
                                /* length changes and truncation.       */
#define ERX_FRMT_A      "A "    /* Data Type A                          */
#define ERX_FRMT_AV     "AV"    /* Data Type AV                         */
#define ERX_FRMT_B      "B "    /* Data Type B                          */
#define ERX_FRMT_BV     "BV"    /* Data Type BV                         */
#define ERX_FRMT_D      "D "    /* Data Type D                          */
#define ERX_FRMT_F4     "F4"    /* Data Type F4                         */
#define ERX_FRMT_F8     "F8"    /* Data Type F8                         */
#define ERX_FRMT_I1     "I1"    /* Data Type I1                         */
#define ERX_FRMT_I2     "I2"    /* Data Type I2                         */
#define ERX_FRMT_I4     "I4"    /* Data Type I4                         */
#define ERX_FRMT_K      "K "    /* Data Type K                          */
#define ERX_FRMT_KV     "KV"    /* Data Type KV                         */
#define ERX_FRMT_L      "L "    /* Data Type L                          */
#define ERX_FRMT_N      "N "    /* Data Type N                          */
#define ERX_FRMT_P      "P "    /* Data Type P                          */
#define ERX_FRMT_T      "T "    /* Data Type T                          */
#define ERX_FRMT_U      "U "    /* Data Type U                          */
#define ERX_FRMT_UV     "UV"    /* Data Type UV                         */

   char                  szErrorText[40];

}  UECB;

The file usrtrpc.c is an example of the SAGTRPC user exit. It is delivered in the Broker user exit directory. See Directories as Used in EntireX.

Using the User Exit Interface Fields

The user exit provides two separate functions, Convert and GetLength. The field eFunction indicates the function to execute.

Errors

Both functions can send an error, using register 15 in the range 1 to 9999 to SAGTRPC together with an error text in the field szErrorText.

  • A value of 0 returned in register 15 means successful response.

  • Error 9999 is reserved for output buffer overflow. See Convert Function.

  • When an error occurs, the conversion of the message will be aborted and the error text will be sent to the receiver (client or server). The error is prefixed with the error class 1011. See Message Class 1011 - User-definable SAGTRPC Conversion Exit.

Example:

The user exit returns 1 in register 15 and the message "Invalid Function" in szErrorText. The receiver gets the error message 10110001 Invalid Function.

Convert Function

This function has to be executed when the contents of eFunction match the definition USRTRPC_FCT_CONVERT.

uReturnedLen must be supplied by SAGTRPC's user-written conversion exit. Its value must be set to the length of the output buffer.

All other interface fields are supplied by the Broker and must not be modified by SAGTRPC's user-written conversion exit.

The incoming data is located in a buffer pointed to by pInputBuffer. uInputLen defines the length.

The outgoing converted message must be written to the buffer pointed to by pOutputBuffer. The field tr_olen defines the maximum length available.

For variable length data such as AV and KV, an output buffer overflow can occur if the message size increases after conversion or the receiver's receive buffer is too small. In this case error 9999 "output buffer overflow" must be returned, which calls the GetLength Function for the remaining fields.

GetLength Function

The GetLength function evaluates the needed length of the output buffer after conversion. An actual conversion must not be performed. The length needed must be returned in the field uOutputLen.

The GetLength function is called for remaining fields after the Convert function returned the error 9999 "output buffer overflow".

The purpose of this function is to evaluate the length needed by the receiver's receive buffer. This length is returned to the receiver in the ACI field RETURN-LENGTH. The receiver can then use the Broker ACI function RECEIVE with the option LAST together with a receive buffer large enough to reread the message.

Character Set and Codepage

The character-set information used is the same as in the user-written translation routine and is taken from scode (for the sender), rcode (for the receiver) and bcode (for the Broker). The character-set information depends on the direction information given in the field eDirection. See the following table:

eDirection From Character Set To Character Set
USRTRPC_DIR_SENDER_TO_BROKER scode bcode
USRTRPC_DIR_SENDER_TO_RECEIVER scode rcode
USRTRPC_DIR_BROKER_TO_RECEIVER bcode rcode

Alternatively, the codepage as derived from the locale string mapping process is provided in uCpSender (sender codepage), uCpReceiver (receiver codepage) and uCpBroker (Broker codepage), and can be used to find the correct conversion table. See the following table and also Locale String Mapping.

eDirection From Codepage To Codepage
USRTRPC_DIR_SENDER_TO_BROKER uCpSender uCpBroker
USRTRPC_DIR_SENDER_TO_RECEIVER uCpSender uCpReceiver
USRTRPC_DIR_BROKER_TO_RECEIVER uCpBroker uCpReceiver

Software AG IDL Data Types to Convert

The field sFormat provides the SAGTRPC user-written conversion exit with the information on the IDL data types to convert. Each data type can be handled independently.

sFormat Data to be converted Notes
FMTA IDL data type A 1, 3, 4
FMTAV IDL data type AV 4, 5
FMTB IDL data type B 1, 2, 7
FMTBV IDL data type BV 1, 2, 7
FMTD IDL data type D 1, 2, 7
FMTF4 IDL data type F4 1, 2, 7
FMTF8 IDL data type F8 1, 2, 7
FMTI1 IDL data type I1 1, 2, 7
FMTI2 IDL data type I2 1, 2, 7
FMTI4 IDL data type I4 1, 2, 7
FMTK IDL data type K 1, 3, 4
FMTKV IDL data type KV 4, 5
FMTL IDL data type L 1, 2, 7
FMTN IDL data type N 1, 2, 7
FMTP IDL data type P 1, 2, 7
FMTT IDL data type T 1, 2, 8
FMTU IDL data type U 1, 2, 7
FMTUV IDL data type UV 1, 2, 7
FMTUSER RPC user data such as user ID, library, program... 1, 3, 4
FMTMETA RPC metadata 1, 2, 7
FMTFB RPC format buffer 1, 2, 7
FMTSB RPC metadata variable length 4, 5, 7
FMTPRE Preview data 4, 6, 7

Notes:

  1. Field length is constant.
  2. The field content length must not increase or decrease during conversion. If this happens, the user exit should produce an error.
  3. If the field content length decreases during the conversion, suitable padding characters (normally blanks) have to be used.
    If the field content length increases during conversion and exceeds the field length, the contents must be truncated or, alternatively, the conversion can be aborted and an error produced.
  4. If the contents are truncated, character boundaries are the responsibility of the user exit. Complete valid characters after conversion have to be guaranteed. This may be a complex task for codepages described under Arabic Shaping, EBCDIC Stateful Codepages or Multibyte or Double-byte Codepages. For single-byte codepages it is simple because the character boundaries are the same as the byte boundaries.
  5. The field length can decrease or increase during the conversion up to the output buffer length. The new field length must be returned in uReturnedLen. If the output buffer in the Convert function is too small, error 9999 must be returned to the caller.
  6. The field buffer should continue to be converted until the output buffer is full or the input buffer has been processed. If the field content length increases or truncations occur, no error should be produced. If the field content length decreases, there should be no padding. The new field length should simply be returned to the caller.
  7. Codepages used for RPC data streams must meet several requirements. See Codepage Requirements for RPC Data Stream Conversions. If these are not met, the codepage cannot be used to convert RPC data streams.

Start of instruction setTo compile and link the SAGTRPC user exit

Configuring SAGTRPC User Exits

The user-written SAGTRPC user exit shared library/object must be accessible to the Broker worker threads.

Start of instruction setTo configure SAGTRPC user exits

  1. Copy the user-written SAGTRPC user exit shared library/object into the EntireX lib directory.

  2. In the Broker attribute file, set the service-specific attribute CONVERSION to the name of your SAGTRPC user exit. Example: 

    CONVERSION=(libmytrans.s[o|l])

    or

  1. Place the user-written translation routine shared library/object in a directory of your choice.

  2. In the Broker attribute file, set the service-specific attribute CONVERSION to the full path name of the directory of the SAGTRPC user exit. Example: 

    CONVERSION=../mydir/mytrans/libmytrans.s[o|l]

Start of instruction setTo configure locale string defaults

  • If the broker's locale string defaults do not match your requirements, we recommend you assign suitable locale string defaults for your country and region. See the appropriate attribute under Codepage-specific Attributes for information on customizing broker's locale string defaults, and also Locale String Mapping.

Start of instruction setTo customize mapping of locale strings