In a recent discussion in the Slide mailing list we were talking about using Lucene as an indexer for Slide. One of Slide's features is full transactional support.
Only evaluating possibilities and not even sure if this is really needed for Slide, but do you people see any way to add transactionality to Lucene? This would mean you can update one or more indices in one transaction and roll back or commit them atomically while concurrently having an isolated access from other transactions.
This may be a track to follow (and maybe not): In Slide there is an implementation for transactional access to a file system. My idea was to make it accessible for Lucene as well. But then, as far as I understand it, Lucene operates on files, not on pure streams, right? This is currently not supported by the the transactional implementation mentioned above (and probably won't be ever).
Addtionally, storing Lucene indices to BLOBS in RDBMS's is no good idea either, I guess, as Lucene makes heavy use of the skip method in input streams. As input streams coming out of JDBC drivers for shure are very slow on this operation (especially a problem when indices grow large) this really does not make sense, does it?
Thanks for any hints in advance,
Oliver
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
