I propose to create a new datastore implementation, closely based on the current one. The new implementation will: - Keep the actual data in the database, not just pointers to it. - Therefore, it should never run into problems with lost blocks. - It will reduce the locking granularity considerably; with the current implementation, we have one big lock for the datastore. With the proposed change locking is implemented by the database. - However, the database will need to move fairly large amounts of data around from time to time, which may result in a reduction in performance.
Any comments? Is this a good idea? It's not a very high priority item, but it seems it might be sensible to do it before 0.7, and it should be fairly easy. -- Matthew J Toseland - toad at amphibian.dyndns.org Freenet Project Official Codemonkey - http://freenetproject.org/ ICTHUS - Nothing is impossible. Our Boss says so. -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 189 bytes Desc: Digital signature URL: <https://emu.freenetproject.org/pipermail/tech/attachments/20060512/73cc70ab/attachment.pgp>
