Jean DOUGNAC
TELIS Télécom
8, rue Sainte Barbe BP2089
F-13203 MARSEILLE CEDEX 01
FRANCE

Tel : +33.91.13.95.11
Fax : +33.91.91.63.43

INTRODUCTION

WHAT A MULTILINGUAL APPLICATION IS

CONSULTING MULTILINGUAL INFORMATION

EXTENTED NOTION OF CHARACTER SET

EXISTING SOFTWARE CHARACTERISTICS

INTRODUCTION

In the scope of the WWW internationalization workshop, this paper is based on the experience consisting to design a multilingual software, to support originately telematics applications on a Worldwide basis.

Similar technical questions apply when developping multilingual applications on the World-Wide Web. This paper introduces the main aspects to consider for that purpose.

WHAT A MULTILINGUAL APPLICATION IS

Typically an application suggesting the end user a list of languages to select his/her preference to work. In the context introduced above, we may imagine a multilingual service presenting first a banner to choose a working language among several various scripts : English, French , Japanese , Arabic, Russian, Hindi or Hebrew in the following screen example.

CONSULTING MULTILINGUAL INFORMATION

Multilingual information is consulted by the mean of a terminal station, whose functional diagram could be the following one.

On the left part of the block diagram, the network interface links the terminal to the server providing the information to consult.

In a multilingual approch, this server must specify to the terminal :

• the character set currently used
• change of current character set when required
• miscellaneous aspects of fonts like character colors, size, underline effect and so (mechanism conform to CEPT2 in the diagram above - to translate in HTML commands).

Selecting and switching character sets are trigerred by ISO 2022 escape sequences.

Unless identifying character sets as private sets (a non recommended method), their identification is standardized by ISO 2375. Keep in mind a big part of confusions in character sets identification is due to the fact :

• several identifiers are assigned to similar languages or scripts (typically latin variants of ISO 646)
• diacritical marks are supported by the mean of incompatible methods (typically in Western Europe stand alone diacritical marks of ISO 6937 versus combined characters of ISO 8859)
• most of Eastern scripts are not yet referenced.
  

EXTENTED NOTION OF CHARACTER SET

Character set standards specify codes assigned to characters, respecting as much as possible the following principles :

• each character of a set has a code and this code is unique, despite this character may have various graphic representations, depending for instance of its position within a word. The case of Arabic letter Aïn (hex code 59 in ISO 9036) is typical
• this assignment conforms to the traditional alphabetic order of characters as much as possible. Sorting words is a problem highly simplified.
  

In addition to code assignment to characters, management of a character set also includes to process :

• the graphic representation of characters
• the arrangement of characters on a keyboard
• the script rules governing typically oriental scripts : graphic representation of a character depending of its position in a word (e.g. Arabic letter Aïn), two or more characters combining in a complex graphic representation (e.g. consonantal clusters in Devanagari), capability to superimposed one or more diacritical marks to a letter (e.g. support of the Korean Hangul in a 7 bit coding environment).
  

Points leading to a representation of a character set by the mean of four planes dealing with:

• code assignment to characters
• graphic representation of characters
• keyboard layout
• script rules management

Considering the management of the three last planes are proprietary, the software introduced at the beginning of this paper was facing the problem to manage a wide range of script rules maintaining the code volume below 640 Ko memory limit of MS/DOS.

The possible implementation was to distinguish :

• a generic and universal module part of the station emulation (called 'Script Processor' on the Terminal Block diagram above)
• the rules specific to a language recorded in addition to the corresponding font file.

In addition to this architecture characteristics, font files also discribe the keyboard layout.

Reconfiguring this layout is a station local function, transparent for the management of remote applications.

Since these mechanisms are proprietary instead of standardized, using such fonts, station installed or downloaded implies to deliver in the same time to the station the portion of code capable to interpret these proprietary data.

It is strongly suggested to consider the implementation of such a portion of code by the mean of an applet.

EXISTING SOFTWARE CHARACTERISTICS

Registered internationally under ORIENT GATEWAY(R) trademark.

• Support of more than 36 scripts worldwide
• Support of multi-bytes character sets (GB2312, JIS X208, KSC 5601)
• Management of conjunct characters, multi-diacritical signs, context analysis
• Combinaison of different text orientations (e.g. left to right & right to left)
• Local reconfiguration of keybord layouts a way transparent for remote applications
• Support of CEPT2 format effectors (parallel mode)
• Character sets management conform to ISO 2022 & ISO 2375
• Support of DRCS & private character sets
• HP PCL4 gateway for printing purpose

Publication

" Terminal multilingue : une approche universelle " (Multilingual terminal : a universal approach)

J.DOUGNAC, Télécom Review, Paris (France), Dec 1995