On Thu, Feb 23, 2006 at 07:48:43PM +0200, Jusa Saari wrote: > On Sat, 18 Feb 2006 01:28:41 +0000, Matthew Toseland wrote: > > > Is it worth breaking backwards compatibility for the 0.7 datastore (with > > prior builds of 0.7) to fix an inherent 64TB limit? > > > > The code uses an int for offsets into files, which is easily fixed. > > However it also uses, on disk, an int for block numbers. This means > > datastores are limited to 2G * 32K = 64TB. Normally I wouldn't regard this > > as a big problem, but since we are in pre-alpha, and since there isn't > > that much content, I'm inclined to make the change... > > If you switch to using longs, you will propably lose more disk space from > having to store extra 32 bits per chunk than gain, since it is unlikely > that anyone has 64TB disk space or would donate it to Freenet even if they > did...
It's a matter of not making life difficult for ourselves later on. It's easy and safe to switch now. It won't be later. > > Come to think of it: would it be possible to use PostgreSQL to store the > datastore, with a plugin perhaps ? Would it offer speed advantages ? We use berkeley DB java edition for the store index. We have a separate file for the actual content. Anyway, this is all irrelevant, as I've done it. :) It *should* be backwards compatible. -- 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/20060301/0dbc7029/attachment.pgp>
