[MUD-Dev] Off-topic! Re: Windows-> Linux TCP/IP stack performance problems

[Adam Wiggins]

On Thu, 10 Sep 1998, Vadim Tkachenko wrote:
> J C Lawrence wrote:
> > non-RFC OS     sends http request.
> > Linux          sends the first segment and waits for an ACK.
> > non-RFC OS     waits 200ms for the seconds segment.
> > non-RFC OS     gives up and sends ACK.
> > Linux          sends more data...
> 
> Probably OT, but happens often enough to mention (I guess, since NT
> 4.0): I've observed a very annoying behavior with Samba when Linux box
> works perfectly as a server and lags enormously as a client (when the
> SMB shares are mounted). It may be related to this problem, too.

If you're using smbmount/smbumount, that's not technically Samba.  smbfs
is a kernel fs driver which happens to overlap quite a bit with the Samba
suite.  I, too, was having quite a few problems with NT 4.0 systems (but
none at all with 95/98) mounted via smbfs; a quick upgrade to the latest
kernel version (I was formerly using 2.0.33, almost a year out of date)
seemed to fix all of them.  It seems that the release of NT 4 'broke' a
lot of the existing stuff.  The actual driver is rife with various special
cases based on the client it's connecting to, since every Windows version
(ie, 95, 98, NT 1, NT 2, NT 3.5, NT 4)  likes to do things in a slightly
different way.

Now it's working like a charm.  If only smbfs used WINS; right now I'm
using a shell script to do a quick smbclient lookup, sed out the IP string
from the smbclient output, and then pipe that into smbmount.

Adam