On Oct 5, 12:09pm, David Gluss wrote:
> NFS expects a reliable protocol like TCP, and PPP supplies a protocol
> that may require retransmit by the higher level layer.  Of course,
> I don't even know for sure that that is true. (about either one)
>
> --David Gluss    david@pure.com     408-524-3028

I don't think that NFS cares about the protocol that is being used.  Also
you're confusing your layers by comparing PPP and TCP.  TCP runs over the top
of IP as does UDP, which is what NFS is built upon (in most cases, mind you).
 PPP rides over the serial line providing a relability in the lower levels (ala
phys and data link).  IP then runs over the top of this and at the top
(basically TCP/IP is a four layer model) is TCP and UDP.  NFS has error control
built in as UDP doesn't provide for ordering or re-transmit.

So suggesting that NFS over PPP isn't working because PPP doesn't have enough
reliability as a protocol doesn't quite make sense, in a strict point of view.
 It may be that NFS is causing taxation on PPP that other applications are not
and the whole IP stack melts down for timeouts and excessive re-transmits.

Steve




-- 
======================================================
Steve Acheson                         (415) 604-6555
Sterling Software                     sma@nas.nasa.gov
NASA Ames
Moffett Field, Ca 94035-1000