Subject: Re: ip reassembly time exceeded?
To: Curt Sampson <cjs@portal.ca>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
List: current-users
Date: 01/27/1997 19:09:43
> area network (this one running Windows 95) was having huge problems
> downloading web pages; it was getting ridiculously slow rates,
> probably due to dropped packets. Setting the MTU above 576 fixed
> this.
> 

>This appears to be some sort of TCP problem. I too ran my PPP link
>with an MTU of 296 bytes for some time, and I noticed that connections
>to and from the routing machine (a Sun 3--the remote end is a
>Livingston Portmaster) were fine, but another machine on my local
>area network (this one running Windows 95) was having huge problems
>downloading web pages; it was getting ridiculously slow rates,
>probably due to dropped packets. Setting the MTU above 576 fixed
>this.

Actually, my take is that this sounds *much* more like an IP
fragmentation or reassembly bug.  A tcpdump trace would tell you
which, and at which end it's on.

Alternatively, is there a third machine on your local net, besides the
Sun3 and the losing Windows95 box?  If so, how does *it* do with TCP
connections to web servers, or other sites on a different net from the
Livingston?

If all else fails, reassembly at the Win95 machine would be my first
suspect. If the PPP link was to the Win95 machine itself, TCP MSS
negotation would make IP reassembly unecessary, at least for TCP
traffic.