On Wed, 27 Dec 2000, Benjamin Coates wrote:
> >From "Mark J. Roberts" <[EMAIL PROTECTED]>
>
> >I don't think my (derivative of Brandon's) MSK proposal can be implemented
> >like the other keytypes, because it is very different. It's not an
> >extension of any existing keytype, and it's not a full-fledged keytype in
> >itself. Furthermore, the efficiency of MSKs relies on caching a hashtable
> >of filename=key pairs, and thus it is better implemented as I have, with a
> >MapHandler you instantiate that manages a cache of any number of
> >MapFiles. You feed it a URI, i.e., "freenet:pigdogjournal//", and it sends
> >back the URI of that file in the map. If the map is not already cached, it
> >requests it and caches it.
>
> Why is there an efficiency problem with just re-requesting the map key? The
> node already does cacheing for all keytypes, and that lets MSKs be "normal"
> keys.
No way. A mapkey could be a hundred kilobytes, and contain 100 images that
the browser is immediately requesting. Yeah, most won't be that big, but I
imagine alphabetically broken-up MP3 indices with mapfiles easily that
big.
Stick the damn thing in memory and flush the cache every 10 minutes.
Shorten that interval if the cache grows too big for comfort. Even a one
minute caching time would be very worthwhile. Things like embedded images
are going to be requested in bursts.
--
Mark Roberts
[EMAIL PROTECTED]
_______________________________________________
Freenet-dev mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/mailman/listinfo/freenet-dev
