• by message type: distinguishes messages by the form of their communication (request for information, response, etc (see Categories of Message / E8);
• by keyword: distinguishes the subject matter of messages, bearing in mind the difficulty of ensuring unambiguous keywords in a conference situation (see Keywording and Numbering / E9);
• by conference structure: distinguishes messages in relation to thematic sections of the conference;
• by language: distinguishes different language versions (all forms of which may or may not be available, depending on the policy and capacity of the translation system); eg. 1023E/1023Ef/1023Es code the same message No. 1023 in the original English and French and Spanish translations).

The editorial team, possibly supported by information from various factions and working groups, may find it convenient to constantly scan and sift through the range of messages. The purpose would be to feed in (and perhaps delete) coded relationships to strengthen and sharpen the substantive focus. This suggests the presence of several other types of record in the database:

• Issue/topic records: which hold references to all messages touching on a topic.
• Working group records: which would hold references tied to a working group; the group might, or might not, cover a whole issue or group of issues.
• Author records: which would hold all references made by a particular author.

Thus a working group on a particular issue might ensure the addition of references to existing messages, or might "reject" references from some messages as irrelevant. Clearly each of these record types could be the subject of alternative hard copy outputs as indexes to the sequence of messages (see also Database Products / C2).

There is an interesting further development to allow different "coding philosophies" to be applied by different coding teams. These might be in competition with each other to offer different insights into the conference process, even selling their overviews to the participants in hard copy form for a price. This might be done with parallel databases: one for the message sequence (and its different language versions); and one each for the alternative coding systems.

Coding can be a time-consuming and complex task. To avoid coding problems arising during the conference:

• coding could be omitted;
• coding could be done at a later stage (possibly after dissemination) if appropriate;
• allow for multiple coding (eg. a text containing both "support" and "opposition" to an earlier message should be coded as both);
• codes could of course be changed subsequently if necessary.

Written by Nadia McLaren, Union of International Associations, with revisions and HTML layout by Robert Pollard, Information Habitat: Where Information Lives.