> However, in my experience the DEQNA is actually very nice.  I haven't
> had any problems with mine.  They just sit in a QBus slot and let the
> VAX talk to the network.  They perform fairly well under heavy loads.

With one exception, that's been my experience too.

That one exception was when we had an NFS mount, a MicroVAX mounting
some disk that lived on a Sun.  Certain files just couldn't be read,
and it was really weird; it correlated with the file's contents.
Attempts would read most of the file (if it was large) and then hang,
retrying and producing "NFS server <sun> not responding, still trying"
complaints.

I eventually tracked it down.  When a large UDP packet is sent, it's
fragmented to fit into Ethernet packets.  For some packet sizes, the
last fragment is really tiny.  The Ethernet spec provides for a minimum
packet size; IIRC this is 60 bytes, and indeed snooping on the Ethernet
traffic reveals that the Sun always padded packets shorter than that.
But packets below (IIRC) 64 bytes, the MicroVAX simply never received.
We had kernel source for the VAX's OS, and as far as I could determine,
the DEQNA just wasn't doing anything with them.

The workaround was to mount with -o rsize=1024,wsize=1024; then, data
packets never got large enough to get fragmented.  And non-fragmented
packets, with the IP and UDP and RPC and NFS headers, were always over
64 bytes.

					der Mouse

			       mouse@rodents.montreal.qc.ca
		     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B