[comp.protocols.nfs] NFS over TCP

raj@hpindwa.cup.hp.com (Rick Jones) (04/30/91)

"I liked the topic so much, I started another basenote"

liberally mangled Victor Kiam quote of the day...


The whole topic of NFS over TCP/IP is an interesting one on it's own.
The basic reasons include better operation over WAN's - either by
avoiding, or better coping with, lost packets.

Everyone seems to think that two nodes on an ethernet wont' have any
problems. Is this really the case? There are several workstations 'out
there' that are able to send data with the ether min timing, that
might be talking with nodes that cannot. I would suggest that as a
*small* but real area of benefit wrt NFS/TCP.

WANs do seem to be the big reason. 

$WAG ON

It would also allow one to 'port' NFS over to a TP implementation, and
thus be a migration strategy for OSI. (Unless they finally get
datagram defined?)

$WAG OFF

Does anyone have NFS and/or UDP modules defined for netsim? Perhaps it
would be interesting to get some 'virtual' data for all this
supposition...

rick jones

hwajin@daahoud.wpd.sgi.com (Hwa-jin Bae) (05/04/91)

>>>>> On 29 Apr 91 17:46:52 GMT, raj@hpindwa.cup.hp.com (Rick Jones) said:
Rick> The whole topic of NFS over TCP/IP is an interesting one on it's own.


check out nfs in 4.3-reno release.

*hwajin

--
hwajin@pei.com
Protocol Engines, Inc.