Good point.  IMHO zero copy receives and sends are nice features for a
networked OS.  How they are implemented (zbufs or something else) is less
important than having the feature.
Kyle

-----Original Message-----
From: Bill Studenmund [mailto:wrstuden@netbsd.org]
Sent: Friday, September 06, 2002 6:47 PM
To: kyle.unice@L-3com.com
Cc: thorpej@wasabisystems.com; david@l8s.co.uk; tech-kern@netbsd.org
Subject: RE: zbufs for NetBSD


On Fri, 6 Sep 2002 kyle.unice@L-3com.com wrote:

> Recieving and transmitting large volumes of data is where zbufs would be a
> help.  NetBSD on large data transmits of memory mapped files does a
minimal
> amount of coping, but recieving large amounts of data is another question.
> Does NetBSD perform the same way when receiving data from the network to a
> memory mapped file?
>
> When recieving data that is measured in megabytes per second, and with
> networks moving to 10 Megabytes (100Mbit/sec) to 100 Megabytes
> (1000Mbits/sec), it would seem that doing a zero copy recieve would be
> advantageous.   Preallocating recieve buffers and managing VM tables would
> seem a faster solution than memcpy when dealing with megabytes/sec data
> speeds.

I think the better question is, what do you want? Do you really want
zbufs, or do you want zero-copy receive and think that zbufs are a
convenient way to do that. Because if a developer doesn't like zbuffs (I
know Jason's the only one who's spoken, but I don't know what other folks
think), that doesn't mean we don't want zero-copy receive. It could just
be that zbuffs aren't a good fit for zcr for us, and we need to do
something else.

Take care,

Bill