Another information retrieval mechanism in Microcosm is called Compute Link [35]. Compute Link, basically uses text contained in an anchor as a key to search for all documents that share a similar vocabulary. The difference between Compute Link and the mechanism we are describing in this report is: compute link does not take in consideration that an anchor can have complete different meanings dependent on the context, so for the same anchor in any node, the destinations will be the same. So, compute links is a powerful way of creating generic links, see section 2. But, since words sometimes can have complete different meaning accordingly to the context they are inserted in, sometimes the meaning of every anchor is important by itself, and so it is the context of the anchor inside the node. When we combine the local keywords attached to the anchor with the global keywords of the node, we are trying to consider the meaning of the anchor by itself, but even more the meaning of the anchor in the context of the node it is located in. So we believe that this mechanism can give better results in some cases, but actually it is a service complementary to the Compute Link mechanism. Using our implementation, for the particular case of finding possible destinations for a generic link, we do not consider the keywords attached to the node. Compute link also does not take synonyms into account, while the thesaurus in the new mechanism is able to identify words that have the same meaning. Also this mechanism can be used for any type of viewer, while Compute Link is valid only for text viewers. The advantage of Compute Link over the mechanism proposed in this report is automation: keywords are searched for in the nodes and correlated automatically during indexing, while in the new mechanism the author has to enter keywords manually. We intend in the future to study alternative ways of attaching keywords to nodes and anchors in order to alleviate the extra burden to the user, since we know that the simple naming of nodes and creation of links are already very demanding [36]. Perhaps the natural evolution would be the utilization of conceptual retrieval [37] [38], where the user expresses the information needed as a concept rather than as a collection of keywords. Examples of conceptual queries could be: Give me documents about the mercury contamination in the Amazon, Give me documents about meetings between Mr. Mason and Mr. Blair in the last month [39]. In Southampton, some initial work in the area of conceptual authoring is being undertaken [40].