> How does UIMA-AS currently do the linking of the network-ID to the CAS? Do you use the CAS instance as the key in a hashmap or something like that?
When the CollectionReader fills a CAS the UIMA-AS client assigns it an ID generated by java.rmi.server.URI.URI() before sending it to the service queue. So even if there is only one CAS in the pool it gets a new ID each time it is reused. I'd like to see this ID become part of the CAS object, set by core UIMA whenever a new CAS is filled by a CR or CM. The other use case for what I think of as a document ID would be the same even if different pipelines process the CAS at different times. I assume it would be set by the CR or CM, not by the framework. Since events can be documented with audio & video as well as with text, I think it's OK to use "document" to refer to the Sofa data in a CAS. By convention we've assumed additional views are alternative representations of the same artifact (although they could be entirely independent) so the document ID could apply to the whole CAS. ~Burn
