Michael Wiktowy <spam at mindless.com> wrote:
> From: Oskar Sandberg <md98-osa at nada.kth.se>
> > Client writers should take note that this means the "DataSource" field
> > is now optional. You can stop setting it to a bogus address when you
> > insert, simply leave it out of the message. People using the java code
> > for clients (Freeloader) should set Core.transience to true and the
> > this will be done automatically. The way I am currently handling this
> > is to tolerate null references in the datastore. Possibly not caching
> > the data at all on the first step from a transient node is another
> > option, I'm not sure which is better.
>
> It seems to me that this is introducing a difference between the first
> hop requests and inserts from traffic sniffers to pick up on.
One thing I suggested before was to introduce a No-Cache: {true|false}
header. If a node receives a message with No-Cache: true, it doesn't cache
the data, but passes the message on with the following alteration: flip
No-Cache to false with probability P, otherwise keep it the same. If it
receives No-Cache: false, it caches the data and passes on the header as
is.
So if you receive a message with No-Cache: true, you don't know if sender
was a client/transient node, or was just passing it on (with probability
1-P). However, the originator is guaranteed one hop without caching and
know that it will start getting cached, on average, within 1/P hops after
that.
theo
_______________________________________________
Freenet-dev mailing list
Freenet-dev at lists.sourceforge.net
http://lists.sourceforge.net/mailman/listinfo/freenet-dev
