The schema was expanded to include a set of ancillary tables describing protocols, formats, event timings, and agents (software, hardware, and human) so that we could document the flow of information through any large system such as the DEIMOS instrument control and data reduction package. The Agents schema permits us to document the source language, revision, authorship, and other attributes special to software agents. (This model of information flow corresponds well with the ``gaming table'' model proposed at ASASS V by Noordam and Deich.) As with Memes, a URI field permits the attachment of detailed documentation to any Agent.
A table of Paths documents the passage of Memes between Agents. The attributes of a Path are the sending and receiving agent, a controlling agent, the Meme ID, and a cluster of attributes describing the transaction: format, protocol, event timing, and elapsed time. Since human beings can be agents and Key Entry is an acceptable protocol, it is possible to document the Memes/Agents toolset itself using this schema.
We use a digraph generation package (see Acknowledgments) to generate graphical representations of information flow. A hierarchy of agents and paths was required in order to generate both simplified and detailed drawings. Accordingly, the schema was adjusted to permit the definition of ``superAgents'' which consist of multiple Paths.
The Agents/Paths database can be used not only to generate functional diagrams, but to assist in diagnostic procedures. It can be used to trace the any FITS keyword value back to the originating or authoritative source; conversely, it can reveal which agents handle a keyword and thus may be affected by changes to the syntax or semantics of that keyword; it can reveal which keywords are handled by an agent, and which ``downstream'' agents may be affected if an agent is disabled or altered.
Currently we are addressing the distinction between syntactic and functional relationships among Memes. We have addressed, for example, the fact that the value of NAXIS controls the appearance of keywords NAXISn (a syntactic relationship). We have now to model those relationships in which a set of FITS keywords acts (in use, rather than in syntactic specification) as a state engine. This is particularly relevant for instrument control systems, in which (for example) a HATCHCLO keyword and a HATCHOPN keyword cannot both be true, but may both be false. The solution to this problem will almost certainly require one additional ancillary table in the schema.