Future Extensibility Requirements

There are three design goals to be considered: visual fidelity, editability, and security.

• Visual fidelity involves the desire for users of two consumers/producers to see visually the same thing. This seems like a simple design goal to meet, but, in practice, is not easy to achieve. The difference lies in the capabilities of an Uprev and Downrev consumer/producer. Typically, an Uprev consumer/producer has been extended to have new base capabilities that are not present in the Downrev consumer/producer. As such, the Downrev client does not have the base primitives necessary to express visually the new capability introduced in the Uprev consumer/producer.Consider the case in which PML 2007 has the capability to highlight text with a given color, while PML 2003 does not. Given the desire to have PML 2007 documents interoperate with PML 2003, it is necessary for PML 2003 to some way to express visually that text highlight. For example, it might insert a picture of the highlighted text instead of inserting the text itself, since PML 2003 does know how to deal with pictures.

• Editability involves the desire for two consumers/producers to be able to edit the same content. Using the highlighted text example from above, despite the fact that PML 2003 and PML 2007 have different capabilities, one would still desire to edit highlighted text a regardless of which version of PML is in use. Again, this becomes difficult when the underlying capabilities of the consumer/producer versions are different.

• Security involves the desire to have multiple representations of the same data synchronized. This desire is referred to as security for the reason that out-of-sync representation can have dire consequences. For example, there might be multiple representations for a sensitive piece of information such as a Social Security number. If this piece of information were edited, it would be critical to keep all alternate representations in sync. What if that information were deleted altogether? If only one representation was deleted but others remained, it would be possible for one to have sensitive information in a document when the intent was to have it deleted.

One solution to try to solve the visual fidelity and editability goals is to have two equivalent representations for the same construct. In the highlighted text example above, a picture of the highlighted text (also called a rasterized version of the highlighted text) is an equivalent representation of the highlighted text itself. One might use the highlighted text representation when the underlying consumer/producer is capable of understanding it; otherwise, the picture version would be used.

Clearly, these design goals compete with each other. While a picture representation of text is capable of capturing full visual fidelity of how extended text looks, obviously that representation doesn't offer the same editability properties of text. One can't manipulate a picture of text nearly as easily as the text itself.

The competing nature of these design characteristics requires that one choose an extensibility construct that offers the best mix of desired characteristics. It will not always be possible to have visual fidelity, editability, and security, at the same time.