On Thu, Feb 21, 2002 at 02:55:02PM -0600, Mark J Roberts wrote:
> Kevin Atkinson:
> > Is that the only thing the current freenet code uses?
> 
> Yes, praise the almighty, yes.
> 
> > Implementation, as it is now.   For example is it "normal" for key 
> > requested to take around a minute.  I notice on my node that loading 
> > even of the most popular page, CoE, CoN takes a minute or so after he has 
> > uploaded the new page for the day.  I try this around 8-9pm EST.
> 
> Terrors defy the contemplation of mortals in that code - 77,000
> lines, 2.32 megabytes... I despise it wholeheartedly.
> 
> Every hop is a miracle, the network is so overloaded. Here's the
> trick: we've got to make sure that nodes are never too overloaded to
> accept and forward new requests. There will always be an insane
> amount of traffic, far too much to gracefully handle, but if we
> can't route correctly then the poor user won't even be able to
> download at the incredible rate of three packets per hour left free
> for his use.
> 
> And I haven't even mentioned my favorite subject, malicious attacks.
> 
> Fuck it. You wouldn't happen to have any ideas about how I might
> skip certain packets to the head of the UDP xmit queue, would you?
> Either userland or kernelspace is fine.
> 
> I'm starting to fear that I'll have to read code. I hate doing that.
Errm, the problem that you are experiencing is not the routing, it's
the bandwidth limiting code not working. Most of the time, anyway.
> 

-- 
The road to Tycho is paved with good intentions

Attachment: msg00648/pgp00000.pgp
Description: PGP signature

Reply via email to