A Proposal

The following then is a starting point for a reference model for data interchange standards.

1. The model might have four layers:
   1. The Atomic Layer represents basic data forms. Things such as UNICODE, ASCII, CCITT Group 3 Fax, PCM, and RGB triplets would fall at this level. Basically, this is the smallest meaningful unit in any system. Below the atomic level, the analysis or manipulation of data would not be meaningful.
   2. The Elemental Layer would provide some format for excapsulating atomic data. TIF, GIF, WAV would provide examples at this level. An elemental form provides minimal additional data, but does provide additional data that allows the atom stream to be interpreted. For example, a stream of RGB triplets that constitutes an image does not specify the number of bits for each sample, or the orders of the samples, or the size of the image. These are all things that might be specified at the elemental level.
   3. The Structural layer would provide two additional aspects of complexity. First, this would be the first level at which multiple types of atomic elements would be allowed. Second, it would be at this level that we would specify the structural relationships between elements. Thus, in SGML, we can include text as well as graphics and specify how they are related. Examples of Standards at this level, moving from simple to complex would include XDR, ASN.1, SGML, ODA, and the EDI framework.
   4. The Application layer would deal with standards that make use of structural layer standards to accomplish some application specific goal. Examples of standards at this level include CALS, TEI, STEP, and specific EDI transaction sets.
2. The model scopes data forms. At the atomic layer the scope definition insures that all appropriate data forms are covered. At higher layers -- structural and application -- standards would reference atomic and elemental standards as a way of defining their data scope. Thus we might be concerned with the following data forms at the atomic level:
   • symbolic
     • text
     • non-text(music, math, logic, etc.)
   • graphic
   • audio
     • raw
     • language
   • image
   • video
3. The model maps representations to tightly coupled operations and standards defined for these operations.. This might involve the standardization of operation categories into groups such as creation, display, storage, and retrieval. This was suggested as one model for defining standards by Spring and Carbo-Bearman. Similar efforts will be required by OLE and CORBA as they mature. In terms of historical precedent, the Graphic Kernel System is a display standard for Computer Graphics Metafile representations of graphics. The Common Command Language is a retrieval standard for text elements. Similarly the Basic Encoding Rules are a transmission rules for data elements defined by ASN.1.
4. Finally, the model needs some clear way to show the relationship between syntactic meta standards such as ASN.1 and SGML and standards derived from them such as CALS, HTML, TEI, and Z39.59 in the case of SGML.