From: Sten Drescher (smd@animal.brooks.af.mil)
Date: 04/05/93


From: smd@animal.brooks.af.mil (Sten Drescher)
Subject: Re: [SUMMARY] SLS Experiences
Date: 5 Apr 1993 11:49:34

On Sun, 28 Mar 1993 23:30:39 GMT, becker@super.org (Donald J. Becker) said:

 Donald> Nntp-Posting-Host: metropolis

 Donald> In article <1p3nekINNsqn@uwm.edu> pegasus@csd4.csd.uwm.edu (David R Hucke) writes:
>As promised, here is the summary of results...
>From @rnd.brooks.af.mil:smd@hrt216.brooks.af.mil Thu Mar 18 15:21:00 1993
>
> Well, I've had some problems with the recent SLS. For one
>thing, they've unnecessarily compiled in the ethernet drivers into the
>a1 kernel, without making any use of them.

 Donald> While it's not yet possible, the plan is to be able to load
 Donald> linux over the net with minimal user input. This requires that
        Great! I was going to mangle the doinstall script to do
that, but I guess I'll wait for a more generalized version.

 Donald> the boot kernel have built-in autoprobing, auto-IRQing
 Donald> ethercard drivers.

>On the machines we use here, that causes a1 to lock up, because we have
>WD8003EB boards that the driver doesn't correctly identify as being
>8-bit, and it tries using it as 16-bit.

 Donald> This problem has been resolved (I think). It would have taken
        I thought I had mentioned that you were working the problem. If
not, my apologies. I thought that I had also given an overall thumbs up
to the whole SLS, but that there were a couple snags along the way
(although while doing it, my attitude wasn't quite as good :( ).

 Donald> weeks longer to discover and fix if Peter hadn't provided a
 Donald> broader (not just larger) user base. The self-selected sample
 Donald> that tests early versions of drivers usually use products
 Donald> biased towards higher performance, and none of them tested the
 Donald> drivers with an older 8003.

 Donald> In return for this one-time problem with Linux, future kernels
 Donald> will likely boot out-of-the-box, and you won't have to specify
 Donald> -DFORCE_8BIT or set up the ethercard memory parameters when you
 Donald> compile your own kernels.

        Unfortunately, because of the boneheads in ops here who sue a
non-standard 0x380 I/O address for many of our cards, the SLS still
won't autodetect the WD straight out of the box, but as long as I can
get it on one system, I should be able to recompile an A1 replacement
with minimal problems.