All verbs currently known to the gossip system.

True if new-item is identical with, or less specific than, known-item.

If we already know ‘Bad Joe killed Sweet Daisy’, there’s no point in learning that ‘someone killed Sweet Daisy’, but there is point in learning ‘someone killed Sweet Daisy with poison’.

True if known-value is the same as new-value, or, for each key present in new-value, has the same value for that key.

The rationale here is that if new-value contains new or different information, it’s worth learning; otherwise, not.

Return a character specification like this character, but comprising only those properties this gossip is interested in.

Return a location specification like this location, but comprising only those elements this gossip is interested in. If none, return nil.

TODO: write docs

Infer a new knowledge item from this item, following this rule.

Integer representation of how interesting this character is to this gossip. TODO: this assumes that characters are passed as keywords, but, as documented above, they probably have to be maps, to allow for degradation.

Integer representation of how interesting this location is to this gossip.

Boolean representation of whether this character is interesting to this gossip.

True if anything about this news item is interesting to this gossip.

True if the location of this news item is interesting to this gossip.

TODO: write docs

Is this verb interesting to this gossip?

True if this news item is already known to this gossip.

This means that the gossip already knows an item which identifiably has the same or more specific values for all the keys of this item except :nth-hand, :confidence and :learned-from.

Return a gossip like this gossip, which has learned this news item if it is of interest to them.

Return a set of knowledge entries that can be inferred from this news item.

Topics of interest to gossip agents. Topics are keyed in this map by their verbs. The keys associated with each topic are the extra pieces of information required to give context to a gossip item. Generally:

• actor is the id of the character who it is reported performed the action;
• other is the id of the character on whom it is reported the action was performed;
• location is the place at which the action was performed;
• object is an object (or possibly list of objects?) relevant to the action;
• price is special to buy/sell, but of significant interest to merchants.

Characters

TODO but note that at most all the receiver can learn about a character from a news item is what the giver knows about that character, degraded by what the receiver finds interesting about them. If we just pass the id here, then either the receiver knows everything in the database about the character, or else the receiver knows nothing at all about the character. Neither is desirable. Further thought needed.

By implication, the character values passed should include all the information the giver knows about the character; that can then be degraded as the receiver stores only that segment which the receiver finds interesting.

Locations

A ‘location’ value is a list comprising at most the x/y coordinate location and the ids of the settlement and region (possibly hierarchically) that contain the location. If the x/y is not local to the home of the receiving agent, they won’t remember it and won’t pass it on; if any of the ids are not interesting So location information will degrade progressively as the item is passed along.

It is assumed that the :home of a character is a location in this sense.

Inferences

If an agent learns that Adam has married Betty, they can infer that Betty has married Adam; if they learn that Charles killed Dorothy, that Dorothy has died. I’m not convinced that my representation of inferences here is ideal.