----- Original Message -----
From: "Brandon" <[email protected]>
To: <freenet-dev at lists.sourceforge.net>
Sent: Thursday, August 03, 2000 11:09 PM
Subject: Re: [Freenet-dev] Datastore formats and scalability


>
> > Heh.  But then, speed may matter if you have large HTLs do to the
> > combined overhead of all the nodes your request is going through.  Of
> > course, all this falls into the category of being a bunch of
> > opinions.  Another good reason for a fast node is that it will take up
> > less CPU time while you play Quake III. ;)
>
> If all nodes ran faster implementations, that might make the network run
> faster. Then again, the combined latency of the network might be
> significant enough that the increased speed of the nodes doesn't make a
> significant difference. Also, the design of the network attempts to
> minimize the number of hops required to get the file. Also, efficiency in
> terms of speed and efficiency in terms of resource usage do not always
> coincide. But of course, optimize away and we'll see what happens.

Would Seripas be able to test what happens when you are running node
software at a given speed?  If not, could that be added?  It might be useful
to find out.

>
> > Well, I might put in compatibility with the config files used by the
> > Java nodes, but I am probably going to have another config file such
> > as /etc/nfreentd.conf which is specifically for nfreenetd.  As for
> > stuff like command line syntax, nfreenetd is meant to be started up
> > with something such as the SVR4 style run level system, not from the
> > normal user command line.  Ultimately, I want nfreenetd to be used
> > like any other Unix network service daemon.
>
> It would be good to support the entire .freenetrc config file options and
> minimize what goes in /etc/nfreenetd.conf.
>
> > As for protocol standards compliance, I will make sure that the
> > *protocol* used by nfreenetd will stay the same, but other behavior
> > such as the criteria for deleting files from the datastore may differ
> > from that in the Java node.  For example, nfreenetd will not have a
> > fixed number of files that it can store data on (even if they are not
> > in the datastore).
>
> This is very arguable, but I think it is *good* to have different
> behaviour for different nodes. Efficiency is maximized by having
> everything work the same, but a network where lots of nodes work
> differently is harder to attack.
>
>
>
> _______________________________________________
> Freenet-dev mailing list
> Freenet-dev at lists.sourceforge.net
> http://lists.sourceforge.net/mailman/listinfo/freenet-dev
>


_______________________________________________
Freenet-dev mailing list
Freenet-dev at lists.sourceforge.net
http://lists.sourceforge.net/mailman/listinfo/freenet-dev

Reply via email to