Re: How does Lachman fit into the history of TCP/IP on Unix?
Author:   Van Jacobson <van@ell.ee.lbl.gov>
Date:     1996/03/23
Forum:    comp.protocols.tcp-ip

In article <4iu6sr$nm7@noao.edu>, W. Richard Stevens <rstevens@noao.edu> wrote:
>As yet another comment on this thread ...  not only are streams not
>the right approach, but the BSD design is not right either, when
>you get into real high speed networking.  Van Jacobson has described
>his prototype (which unfortunately has never made it outside his lab)
>which changes the framework completely.

There are some more detailed design notes in the slides from
a four hour class I taught at Berkeley in 1992 available at
ftp://ftp.ee.lbl.gov/talks/vjnkarch.ps.Z.  I attempted to start
at the send syscall & go down to the driver then back up the
receive path, showing all the code along the way.  I used UDP
instead of TCP for the example since I've talked about the TCP
architecture before and, besides, UDP made all the code fit on
one slide.  I'm afraid the notes won't make a whole lot of sense
without my accompanying babbling (they were mostly to remind me
what I wanted to talk about) but they have a bit more info than
the other examples Rich mentioned.

 - Van