On Mon, Jul 12, 2004 at 11:10:01AM +0000, Wayne McDougall wrote:
> There's lots of cool stuff with averaging limits, and immediate limits,
> and gradual adjustment. Together with incoming being not directly under
> control. It works very well for those of us with monthly bandwidth caps.

It does?! I thought the average limiter didn't work...
> It is my opinion that a node works [much!] better if it doesn't have the 
> inputBandwidthLimit set at 0, but at a realistic value. That is based on
> month long tests but only on my own (128Kbit) node. From the little I can
> pick up as to how cooperative bandwidth limiting might work it makes sense
> to me theoretically as well.

Hmmm, that sounds rather strange, as I'm pretty sure the input limiter
doesn't work... and ANY limiting will increase latencies significantly.
Limiting output will normally have a knock-on effect on input, unless
you are downloading lots of files locally... of course if you limit
input to the same as output, you won't give away that fact so easily :).
> So if it was my node I'd have:
> inputBandwidthLimit=24000
> outputBandwidthLimit=24000
> (and if I was going away for a weekend or more I'd crank them both up to
> 48000 if no one else was using the bandwidth).
> > I'm not highly motivated right now to update the Java environment.   So far 
> I haven't
> > had observable environment errors.  The security issues I'm aware of involve
> > violations of the security sandbox - a moot point with freenet - and a JVM 
> crash/Dos,
> > which I'll deal with when I see crashes.  If you're aware of something more 
> serious,
> > please tell.
> Nope. I'd agree with all your comments. 
> >  > What does FRED have to say for itself?
> >  >
> >  >
> > 
> > Wow, lots of pretty graphs .  The numbers at the top of the report:
> Very pretty. They don't mean much to me so I go for the "Classic" look
> of Connections, and More Details if I'm browsing.
> > Connections open (Inbound/Outbound/Limit) 198 (132/66/200)
> > Transfers active (Transmit/Receiving)     24 (13/11)
> > Data waiting to be transferred            1,285 Bytes
> > Total amount of data transferred          4,483 MiB
> Perfect. That's exactly what I'd expect to see after say 2 days uptime?

> >  > My personal experience (counts for very little) is that it took 9 days to
> >  > become better connected - then suddenyl everything started working
> >  > beautifully.
> > 
> > Double plus thank you!  I can wait a couple weeks.  I saw the claim that 
> freenet
> > could be competetive with bittorrent, and was worried that I'd botched 
> something

It can be, for large popular files. Once they get started. If you use
enough threads. OTOH, for smaller files, e.g. fproxy, latency is
generally quite high.

> > badly.  I think I've been through about four of the FAQ pages, a couple of 
> which
> > have a subtextual hint that it may be quite a while before one's node is 
> fully
> > connected, but not much idea of the scale of "quite a while".  Setting 
> expectations
> > is important.

Nine days is ridiculous. We must do something about it. :(

> Bittorrent rocks. But it will always max out my connection.
> Freenet easily outperformed Shareaza/Kazaa in my one test. BUT..a big BUT...
> this was a movie file that CofE mentioned (and linked to) in his flog as
> a file he downloaded as a test. I'm guessing there would be many people
> like me who also downloaded the file as a test. Which would mean that Freenet,
> if operating as designed, would replicate more and more of this data
> throughout the network (a reverse Slashdot effect). That would certainly be
> consistent with my observations.
> And just to expose my complete Freebieness (a freenet newbie and I've only
> recently picked up that term recently), I had always done my downloading
> through the built in FRED interface. Ok, nice for built in, but now I do all
> my (few) downloads through Fuqid. What a difference. Haven't looked at
> anything equivalent for Linux.

Okay, what's the main advantage? Maybe we can improve the fproxy
> My interest is websites that can never get slashdotted and can host large
> files while sharing the load, rather than file-sharing...

Yeah, that would be cool, if it really worked, and if we had enough
hosts to be able to worry about such things!
Matthew J Toseland - [EMAIL PROTECTED]
Freenet Project Official Codemonkey - http://freenetproject.org/
ICTHUS - Nothing is impossible. Our Boss says so.

Attachment: signature.asc
Description: Digital signature

Support mailing list
Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support

Reply via email to