*BSD News Article 18639


Return to BSD News archive

Newsgroups: comp.os.386bsd.misc
Path: sserve!newshost.anu.edu.au!munnari.oz.au!constellation!rex!ben
From: ben@rex.uokhsc.edu (Benjamin Z. Goldsteen)
Subject: SUMMARY: NetBSD questions
Message-ID: <CAGF86.3qn@rex.uokhsc.edu>
Date: Tue, 20 Jul 1993 08:21:42 GMT
Reply-To: benjamin-goldsteen@uokhsc.edu
Organization: Health Sciences Center, University of Oklahoma
Lines: 216

Here is a summary to the questions I posed a bit ago on this newsgroup:
[Stuff in "'s was what I got in e-mail]

> -Why was tip/cu essentially dropped?  Is it really expected that I
> download everything onto DOS floppies and then read them back using
> mread?  I hesitate bring my PC into work to connect it to the network --
> using the modem automates the process (I do it over night)?  To get it
> to work, I had to create /var, /var/spool, /var/spool/lock (or was it
> uucp?) and copy "tip" to "cu" ( because there was no "ln").  To figure
> out what I had to do, I had to extract base08 on my main UNIX machine
> and browse through the man pages. 

"tip is on the set of install floppies, but i couldn't test it, so i
didn't document it..."-chris

It actually works fairly good -- it just needs to be documented,
include the appropriate directories, etc (and perhaps include the ln of
tip to cu).

> -After I was able to get the modem to function under NetBSD, I had a few
> problems with rz.  Is there a problem with using a newer version of RZ? 
> This one lacks crash recovery and multiple file receives (the latter can
> be handled by a tedious shell loop). 

"Yes.  You need a modified version to work with tip - see on netcom.com:
~ftp/pub/alm/rzsz-3.24.tar.Z"  Others mentioned that Chuck Foresberg
doped Tip support in his freeware product, limited licensing policies,
etc.  Unfortunate...I have thought many times about writing a
BSD-type-free Zmodem program that worked, but never got around to it... 

> -Why do I keep getting SILO overflows?  I have a 386SX-20 with 16450
> UARTS connected to an external Zoom VFX V.32bis/V.42bis modem.  I have
> to use software flow control due to a deficiency in our terminal server. 
> This setup generally works at 9600 baud (which I could not say for
> 386BSD-0.1), but occasionally it stalls.

    This one sounds like it will be fixed -- there are faster drivers
around -- they just are not bundled, yet.  I have heard of a few around
-- I just wished they were bundled. 

"I ignore the silo overflows and never had a problem.  Ironically I
quit having silo overflows after I got a faster modem V.32 and set the
interface speed to 38,400.  This works much better than my 2400 bps
modem at 19,200."

> -Are there any plans for virtual screens (ala Linux)?  I found them
> pretty handy when I tried Linux out a while back.  One thing about Linux
> is that the terminal drivers feels much faster.  Most UNIX consoles have
> pretty slow consoles (they expect you to scoot right on into X), but
> Linux felt like direct writes to video RAM.

"Try codrv
This is a console driver - I am not sure if NetBSD has ported it.
It is in the 0.2.4 patch kit (for 386BSD).  It will be in 386BSD 0.1.5
(to be alpha released this week??)"

> -Some commands seem to me missing -- like mfs.  I wanted to try out
> memory file systems, but I could not seem to get anything going.  "mfs"
> does not seem to exist -- the way I understand the man pages, I create a
> memory-block device using mfs and than mount it.  "adduser" does not
> exist either.  Some man pages seem to missing, too. 

I was told I could "ln newfs ms".  I was also told I had to make a dummy
filesystem, but it would not have to be as big as the MFS (I think it
does have exist because I tried using a nonexistent name).

"if you want to use an MFS, put a line like:
/dev/sd0b               /tmp                    mfs rw,-s=32768 0 0
in your /etc/fstab."

'you shouldn't use "mfs" to create an MFS, but rather "mount_mfs"'

Conclusion: this program does not seemed to be used very often...

    I was also told I could edit the password file manually.  That was a
general comment.  There are two ways I rate a system -- does it confuse
me or mislead me at first -- and once I learn its peculiarities, does
it work.  Linux (for me at least) failed the last one.  Missing adduser
is just an annoyance (well, I was not sure about the new
/etc/passwd:/etc/passwd.master format, too).  I am fairly well versed in
UNIX so I was not completely lost...  

> -Is there anyway to make it so that I can make my system think its on a
> network even if I do not have an Ethernet card?  I would like to route
> traffic such that if I were to do "ping atom" (atom is my machine), it
> would go through the loopback device (and making atom an alias for the
> loopback address is not what I mean).  I would also like to try out the
> NFS through the loopback, too.

"the next version of NetBSD will have support for multiple loopback
interfaces.  you can do this with them." -chris 

"# ifconfig lo0 inet localhost
Works for me."

> -Is there any plans to ship various programs like Top or Joe with
> NetBSD?  Linux SLS did that and I thought it was pretty good idea. 
> In order to keep things separate, perhaps somebody could just direct me
> to an FTP site of pre-compiled packages?  I usually wish to compile
> packages myself, but I am not sure if I will have the space.  I would
> like to use NetBSD to orient myself with kernel programming, although, I
> probably need a system upgrade (at least more HD space).

    I was essentially told no, but that I could ftp to
alpha.gnu.ai.mit.edu.  BTW, Joe's scrolling is not slow (when you do
stty 38400 or setenv BAUD 38400) and it supports multiple windows.  The
new version does have an odd flicker, though.

> > -How do I change the system's timezone?  Do I just move the symbolic 
> link in /etc to point to the midwest rather than the pacific?  On 
> UNIX systems at work, the timezone is an environment variable -- which 
> is preset in /etc/TIMEZONE and /etc/TIMEZONE.csh. 

    One has to rebuild the kernel and change the link

> -Which Ethernet card is the best?  I have heard good things about the
> NE2000 in general, but it sounds like 386BSD/NetBSD's NE2000 device
> driver has some problems.  Does anybody plan to fix that or is there a
> better card?

    On this one I generally heard that "Good" Ethernet cards were:
WD8003S
WD8003E
WD8013EBT
WD8003EP
WD8013EB
WD8003EP
WD8013EBP
WD8013EPC
3c503

because they use the if_ed driver.  There was a caveat about old WD8003
cards not working, but that may be fixed by now.  Does anybody know if a
3COM 3C509 would work (the Etherlink III I think)?  It is supposed to be
one the fastest PC Ethernet cards out there.  We have 3 unused Cabletron
Ethernet cards, one 3C503 in a public PC, and one 3C509 in a staff
computer.  I could probably switch the 3C503 with the Cabletron (thought
it might be a lot of work because the public PC in question is an old
Zenith PC with odd memory mapping).  

Chris said: "buy a wd8013 (SMC Elite-16).  i think that's probably your
best bet." 

Another person, however, said that their NE2000 worked great with
386BSD, but the SMC Elite did not.

> -How does NetBSD-0.8 differ from BSD4.4's networking and filesystem
> code?  BSD4.4 has been available in alpha and beta for some time now --
> has any of it been borrowed for the 386BSD project?  If not, when (if)
> BSD4.4-lite becomes available, would upgrading those subsystems improve
> performance a great deal?  That is, would 386BSD require less CPU power
> to push full Ethernet through TCP/IP or get greater throughput on the
> filesystem by switching?

    It seems 386BSD is already using Mach's (2.5?) VM system.  One
person thought that 386BSD used the BSD4.2 UFS.

On to another question:
My understanding of the various PC-BSD's available are:

386BSD 0.1 - Basic starting point of 386BSD.  Not very robust
386BSD 0.1 + patchkit (currently at 0.2.4) - 386BSD, but it works
386BSD 0.1.5 - Stable version of patched 386BSD
386BSD 0.2 - "The greatest BSD ever made - promised by WJ"
NetBSD 0.8 - offshoot of 386BSD 0.1 + 0.1.5 patchkit designed to be more
  reliable and not require getting 386BSD 0.1 and applying a patch kit
  and recompiling.
NetBSD current - NetBSD that is not binary compatible with 386BSD

"NetBSD also tries to be more portable -- and whereas 386BSD 0.2 goes in
new directions, NetBSD is more like 4.4BSD.  I think 0.2 is the right
way to go, if it can muster enough support."

>> >-Is there any plans to ship various programs like Top or Joe with
>> >NetBSD?  Linux SLS did that and I thought it was pretty good idea. 
>> 
>> This is more the domain of 386BSD 0.1.5 (aka FreeBSD).

"You can get the 386BSD 0.1 0.2.4 (sic) patch kit from freefall.cdrom.com.
The 386BSD 0.1 bin and src distributions are on agate.berkeley.edu (if
not on freefall)

If you are interested in kernel hacking, NetBSD is the right one.
386BSD 0.1.5 trys to  be more stable as it will be distributed on
CD-ROM.  An alpha release should be put up for ftp this week on
freefall.cdrom.com.

One problem with NetBSD current (as opposed to 0.8) is that the kernel
has changed so much, it does not work with many existing drivers.
It is not even binary-compatible with 386BSD now."

Anybody want to clarify any of that?  BTW, do we really need this many
versions?  We seem to need a reliable base version; an extended,
reliable, friendly version based on the reliable base version; and a
research version -- based loosely on the reliable version.  Right now,
we have four released versions, a few available (alpha/beta) versions,
and a few promised versions.  

    One thing I would like to see are more binary too releases -- some
of the bugs fixed in the patchkits are prevent me from aquiring them (I
was never able to download 386BSD 0.1 because the serial driver in the
original did not work -- I never got around to feeding my computer
floppies for an afternoon).  

Thanks for the all the help from:
Bruce Evans <bde@kralizec.zeta.org.au>
Bruce Jackson <jackson@ponder.csci.unt.edu>
Andrew Moore <alm@netcom.com>
Alex Tang <altitude@css.itd.umich.edu>
chris <cgd@postgres.Berkeley.EDU>
Vax <vax@ccwf.cc.utexas.edu>

[I hope I did miss anybody...]
[P.S.I hope nobody minded that I did not attribute responses to their authors]
-- 
Benjamin Z. Goldsteen