On 11/13/06, toad <toad at amphibian.dyndns.org> wrote:
> What do we want the default heuristics to be for containers?
>
> Presently, we have:
> Any file below 64kB is considered for inclusion in a container.
> Only one container will be inserted per freesite.
> It is allowed to grow up to 2MB, then files are rejected and inserted
> separately.
>
> Obviously this sucks:
> - An activelink will pull the whole 2MB container. This means indexes
>   can take a very long time to load!
> - Unused sites are supposed to fall out of freenet; activelinks subvert
>   this, but maybe they should only fetch part of the site, especially
>   given the first point.
> - Larger sites could benefit from more files being included in more
>   containers. And the current rules impose an arbitrary cutoff.
> - Larger files may sometimes be profitably included in the archive.
>
> Proposed solution:
> - Any freesite which fits entirely into a container of 256kB will be
>   inserted as a single container.
> - For larger sites, if the activelink (indicated by a flag) plus the
>   HTML plus the CSS fits into a container of 512kB, it should be
>   inserted as a single container. The rest of the site may be inserted
>   as a single container (excluding very large files - how large?).
> - For even larger sites, we try to divide it up mostly by subdirectory,
>   with a limit of 512kB per container.
>
> Obviously these heuristics can be overridden by clients, which can
> specify which container each file goes into. It has also been suggested
> that we offer a choice of heuristics to clients (by type, by directory
> etc).
>
> Benefits:
> - Small freesites retain the advantages of being a single container. And
>   benefit relative to the current situation because larger files can be
>   included as long as the overall site is within the limit.
> - Large freesites benefit because they can still be containerised (this
>   can be a huge benefit).
> - Indexes load faster, and the activelink only preloads the html for
>   most sites (so the images will fall off if they are not accessed).
>
> Ideas? Details? Objections?

I'd like the proposed size limits doubled but apart from that it sounds good.

>
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.5 (GNU/Linux)
>
> iD8DBQFFWH7vA9rUluQ9pFARAlZEAJ4vgeSTrEPMDKNvy9yPW2nfvDDv/QCeI2fY
> tt8boSX7ozS6otR3ob1dpyY=
> =fFUq
> -----END PGP SIGNATURE-----
>
>
> _______________________________________________
> Devl mailing list
> Devl at freenetproject.org
> http://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
>
>

Reply via email to