Most of the time I expect the node to decide what goes into a container. How can it decide this? I propose the following rules: - Anything which is over 3 blocks compressed (10 blocks uncompressed?) should not go in a manifest. - Anything much larger than the average probably shouldn't go in a manifest. - Content of the same type should go in the same container, in order to optimize compression. - Small uncompressible content should go into a container, but not the same manifest as highly compressible content e.g. HTML.
Hence, the Hardware Book might have two tar.bz2's, one containing all the HTML, and the other containing everything else. The base redirect block would contain the top level metadata for both manifests. Is it useful to split a site up into multiple containers? Certainly we should let the client provide hints as to how to slice up the site, but it may be useful to split it up without reference to the directory structure of the site... Maybe on ClientPutComplexDir, an optional ContainerName for each file? -- 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/devl/attachments/20060519/57f7b01c/attachment.pgp>
