> I notice that the supported hardware section of the INSTALL document
> lists the Intel EtherExpress Pro 10/100B as untested... well, I've got
> one in my PC164, and it's more or less working:

> However, if I send a lot of data from the alpha to another machine on my
> LAN (e.g. ftp a file from the alpha, or run "ttcp -t -s"), I get "fxp0:
> device timeout" messages on the console. Receiving seems to be fine...
> I'm running at 10Mbps, half-duplex.

I've put a lot of packets through 82558s at 100 megabits - both full 
and half duplex.  I'm using a point-point link between a pair of 
Miatas. 

I've seen the timeout messages but haven't investigated.  (The console 
is downstairs where I don't catch things out of the corner of my 
eye.) 

Packets get lost or delayed under heavy or complicated load.  I don't 
have a simple test case.  

A simple UDP request-response test works pretty well.  The TCP version 
gets into a lot of trouble - mostly when the message lengths are 
several MTUs.  I assume the acks are coming back at the "right" time. 

I've run similar tests on PCs/Deskpros.  The PCs seem to work a lot 
better. 
 
If somebody wants to investigate I'll try to localize a test case 
and/or see if my problems correlate with the console timeout msg.