On Saturday 25 January 2003 15:07, you wrote:
> > 1) This doesn't belong in Fred's JVM. We have enough problems
> > undertanding/bounding fred's resource consumption as it is.
>
> I agree - that is what plugin architectures are for.
What do you mean?

When you add a servlet to the mainport configuration setting you are running 
it in fred's JVM.  

Where else are people plugging things in?

>
> > 2) How do you plan to address QOS?  I have asked this question several
> > times and each time it is ignored.
> > If you really want to do streaming you need to have a reasonable QOS
> > gaurantee.   I don't see how you are going to get this from fred.
>
> Increasingly, streaming protocols operate over TCP - what QOS guarantee
> does TCP offer? 
At least the lower bounds can be charcterized.  Perhaps you could help do the 
same for freenet.  See below.

>  Such streaming protocols address this using buffering,
> there is no reason why we can't do the same.
The latencies in Freenet are orders of magnitude larger.

I would be interested in a principled analysis of what it would take to make 
this really work.  How much buffering?  What minimum average rate?

>
> Again, you have no more evidence.
Well I have never seen 30k/sec sustained average transfer rate for a recently 
inserted SplitFiles.

> that it won't work than I do that it
> will. 

> Why not encourage someone to try it so that we can all find out?
> Nothing ventured, nothing gained.
I don't think I could stop fish if I wanted to, which I don't.  Pointing out 
that the Freenet architecture provides no QOS, is not what I would 
characterize as discouragement.  It's more like technical due diligence.

> Why not encourage people to do things which have a reason
>  Personally I am optimistic, with
> Splitfiles I have been seeing overall download rates of around 30k/sec,
> which is more than enough for a FM quality ogg stream.
Which SplitFiles?  What tests have you run?

That is about 5 to 10 times the speed I see for non-established SplitFiles.

And higher than the average transfer I see even for most established ones.

Why don't you organize some tests of SplitFIle retrieval rates?  That way the 
discussion would be less anecdotal.

I am hindered in my testing by the lack of publically available "legitimate" 
-- not porn, not copyright infringing -- SplitFiles.  Having a bunch of small 
known (1 - 64M) SplitFiles in the network would be really useful.

If you (or anyone else) wants to insert insert some stuff, use 
freenet.client.cli.Main.  It's the only client I am aware of that supports 
checksumming.  Don't use fproxy.  It doesn't use the new AutoRequester 
SplitFile insertion code yet.

--gj

>
> Ian.

_______________________________________________
devl mailing list
devl at freenetproject.org
http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl

Reply via email to