Details
-
Improvement
-
Status: Open
-
Major
-
Resolution: Unresolved
-
ManifoldCF 1.6.1
-
None
Description
The framework currently handles CHAINED models by expecting that reference extraction take place on all documents with references, regardless of whether those documents have changed. But a far more common kind of situation is when a document's references are accurately captured by its version information. It would be great to perform incremental crawls in this case without doing any additional reference extraction.
Such a scan would require requeuing of all existing documents, which would make minimal scans have to process more documents. But this could be a positive tradeoff if versioning a document was cheap compared to processing it.
Maybe a new class of models, e.g. VERSIONED_CHAINED, might make sense to consider adding.