In a direct-manipulation system, index commands cannot be entered directly in the document under manipulation. A possible solution is to put them in ``shadow pages'' instead of the document output representation. A shadow document is the original document plus special tags that, among other things, mark logical objects like comments, bibliographical citations, cross references, indexes, etc. These tags are essential to document composition but do not correspond to physical appearance in their original forms. Upon request, the corresponding markers of these tags can be displayed along with the original document for editing purposes. For the user's visual cue, each type of tags can be represented by a different marker symbol. Normally for each tag entered in the document, an embedded annotation can be specified. An additional window can be created to show the associated annotation if necessary. This shadow document approach is widely adopted by direct-manipulation document development or desktop publishing systems such as Xerox Star [14], FrameMaker [15], MicroSoft Word [16], and Ventura Publisher [17].
The primary issue in step I for both paradigms is whether or not a systematic mechanism can be derived for the entering of index commands or tags. Details of a general model that accomplishes this task are given below.