The Next Topics for WWW Internationalization

This is a position paper for the Workshop on Internationalization at the 5th WWW conference in Paris on May 6th, 1996.

About the Author

Martin J. Dürst is one of the coauthors of the internet draft for the internatinalization of HTML. His main research fields and interests are worldwide text processing and multimedia, in particular object-oriented architectures for multilingual/multiscript software and computer-supported font design. He is currently a senior research associate at the MultiMedia Laboratory, Deptartment of Computer Science, University of Zurich, Switzerland. He obtained his PhD from the University of Tokyo, Japan, in Computer Science, and his Masters from the University of Zurich, Switzerland, in Compter Science, Business Administration, and Japanese Studies.

Main Issues in Full WWW Internationalization

With respect to the full internationalization of WWW, i.e. the ability to use the WWW with any language and script currently used worldwide, the following following is what I currently am thinking about and/or working on:

• Issues and Non-Issues in Web Internationalization
• Internationalization of URLs
• Multilingual Typography
• Annotations for Ideographic Texts: Rubi

Current Issues and Non-Issues in Web Internationalization

Most basic issues are solved or close to a solution. The I18N internet draft for HTML, together with related documents, provides these solutions.

In some cases, multiple solutions are proposed to address the needs of a wide range of user communities, software architectures, and organizational structures as well as for backwards compatibility and smooth deployment. As example for this is the various ways in which, both in the HTTP protocol and with hooks in HTML, the encoding of a document on the wire (MIME "charset" parameter) can be specified.

In other cases, a single attribute, such as LANG, provides information for a wide variety of possible services to the user, from sophisticated search services to high-quality typographic rendering.

The abovementionned draft basically addresses the requirements for markup common to, and necessary for, all kinds of multilingual documents or documents written in single scripts. As far as the knowledge of the authors of the abovementionned draft goes, the proposed solutions are very feasible. Of course, they are still very open to suggestions both in terms of content and presentation.

The draft does not address issues that can and should be solved purely on the server side (e.g. MIME "charset" identification and conversion) or on the browser side (e.g. sophisticated rendering and font issues). Also, issues for which special markup may seem desirable for a certain group of multilingual documents, but which can be implemented by the current facilities in HTML, are not discussed. The most discussed case here is allignement of different translations of the same document.

Internationalization of URLs

This is at the moment the most important issue not yet solved with regards to the internationalization of the internet. The issue is highly controversial, with many interest groups involved, most of which have no big interest in internationalization, and with many possible solutions, all of which have advantages and disadvantages. Some of my thoughts on this issue can be found at in a separate document.

Multilingual Typography

This section summarises thoughts and experiences I have made on the subject of multilingual typography. It should be helpful for people designing browsers or multilingual documents.

• Typographic conventions such as the mainly used typefaces, characters for quotations, page layout, ligatures, and so on differ even among languages and local typographic traditions that are considered very close from a character set viewpoint (e.g. French, German, and Italian). For example, French frequently uses ligatures for "st" and "ct", whereas Italian mostly uses no ligatures at all. Also, in German typefaces, upper-case letters are usually smaller than in English or French versions of the same typeface, to account for the higher frequency of upper-case letters in German. For less related languages, scripts, and places, typographic conventions differ even more widely.
• On the web, the typographic quality required and/or delivered varies widely. In some situations, especially if resoures are scarse, bare readability is most important. In other cases, higher quality and better control over the output are required.
• Typography is not a theoretical science, but is a field that is developped with practical examples. It addapts to the needs of the writer and reader and to the technologies and the general aestetic sense of the people at a given time and place.
• There are quite some established examples for bilingual/biscript typography for special works, i.e. dictionaries and scientific texts with examples in a foreign language. There are, however, only very few examples with more than two languages or scripts, and almost no examples with an even wider variety of languages and scripts. Therefore, it can be said that multilingual typography is not currently existing, and has yet to be developped.
• Very important is to recognize that multilingual typography cannot be developped transitively, i.e. if a combination of font A for script AA and font B for script BB work well together, and in addition a combination of font B for script BB and font C for script CC work well together, it cannot be deduced that font A for script AA and font C for script, or for that matter all of the three, will look nice on the same page. This applies to almost every aspect of typography, but in particular to typesize, typeface, and baseline adjustment.
• The context of the language mixture and the ratio of the different languages can be very important for good solutions. In some cases, pieces of another language mixed into the main language of a document should appear as general text, and therefore most of the typographic conventions of the "host" language should be applied also to the "guest" language (e.g. foreign names in a newspaper article). In other cases, the mixed-in guest language pieces should be very well visible and clearly distinguished, for which in most cases special typographic conventions are used (e.g. have a look at your favorite dictionary).

In conclusion, the subject of multilingual typography is barely existing and therefore not very well understood. Good solutions for true multilingual typography can only be developped in connection with the production of true multilingual texts, and it will take quite some time (as for any other aspect of typography) for new conventions to be widely accepted.

Annotation of Ideographic Texts: Rubi

In my view, this is probably the one single item that should be added to the i18n draft. While bidirectional formatting control is important for languages such as Hebrew and Arabic, and superscripts are needed for the adequate display of ordinal numbers in French and other languages, rubi are required in many cases to display ideographic text, especially Japanese, in a fashion accessible to a wide range of readers. My proposal for adding rubi syntax to HTML has the advantage that a good result is achieved even on browsers that don't support this feature. Any and all comments are wellcome.