Text layout systems, whether they are part of word processors, web browsers, or provided as part of an operating system or application framework, now have to deal with complex scripts. They must solve problems such as diacritic placement, line breaking in the absence of word spaces, bidirectional layout, ligature building, and caret placement. The advent of Unicode has meant that the text itself, whatever languages it may contain, now has a common representation; one text layout system can and should handle any content. This paper discusses data representations and algorithms for universal multi-lingual text layout from a critical and comparative stance, based on the author's experience in writing such systems, and makes concrete recommendations for building software that is reliable, fast enough to be used on relatively slow platforms like palm-tops and mobile phones as well as desk-top computers, and as simple as possible - but no simpler.

This paper is aimed at a technical audience. Some programming experience is advisable.