David Brownlee <abs@anim.dreamworks.com> wrote:
> This maillist _is_ 'port-vax', for (I quote)
> "Discussion of issues specific to NetBSD/vax", which explains
> most of 'this group's interest.
   
   A lot of people (me included) have joined this list (and started looking
at NetBSD/vax at all) not because of a great love for NetBSD but because of
the lack of BabyVAX support in other VAX OSes. Therefore, there is nothing
wrong with discussing alternatives to NetBSD on this list. Also this list
is an excellent place to discuss VAX hardware even for an I-don't-do-UNIX
VMSer.
   
> > [adding 4.4 vm system to 4.3 ]
     ^^^^^^^^^^^^^^^^^^^^^^^^^^^
> >         The result would be NetBSD/vax with all its problems [...]
   
   Actually that was about taking the VAX code out of 4.3BSD and porting it
to a mostly 4.4BSD base (this is essentially what NetBSD/vax does), not
about taking the VM from 4.4BSD and porting it to a mostly 4.3BSD base. The
latter would be rather pointless, as it wouldn't bring any new
functionality (except perhaps mmap(2)).
   
> > [...] lack of
> > satisfactory support for anything except MicroVAX II and MicroVAX 3
> > [...]
   
   Note the word "satisfactory".
   
> Ahem, have you _read_ http://www.netbsd.org/Ports/vax/index.html ?
   
   Many, many times.
   
> The 11/750, 11/780, 11/785, and 8600 (the machines on which Ragge
> produced the initial NetBSD/vax port), are if anything the
> original 'true' vaxes, and are certianly not 'MicroVAX II and
> MicroVAX 3'.
   
   Have you ever spent even one minute pondering over how would one go
about actually INSTALLING NetBSD on any of these machines? NetBSD/vax
doesn't support any of the classical (ostensibly non-TMSCP) UNIBUS and
MASSBUS tapes, so unless you are lucky to have a UNIBUS KLESI or a TUK50
(UNIBUS TMSCP controllers), you're hosed. You can only cheat in one of two
ways. First, if you have a copy of the appropriate version of Ultrix for
your VAX (the TK50 and CD-ROM distributions for MicroVAXen certainly won't
do even if you have a drive to read them), you can install it and then use
to install NetBSD. Second, if your VAX had a UDA50 and you also have a Q-
bus MicroVAX with a KDA50, you can attach an RAxx disk to the MicroVAX,
completely prepare it there, then attach it to the VAX and boot from it.
Both approaches have prerequisites that a lot of people can't meet.
   
> Sorry to debunk that statement, but a lot of commercial
> organisations certainly think FreeBSD is stable enough to handle
> their livelyhood:
   
   Back when I still believed in the idea of UNIX on IBM PC, I was running
FreeBSD (v2.2.2 IIRC) and I DID have problems. First, I wasn't able to get
MFS to work. Second, console terminal emulation was quite poor (with both
sc and vt drivers). In the case of vt, it was supposed to emulate VT220,
but it didn't, so I would classify this as a bug, rather than a missing
feature. Third, I wasn't able to make FreeBSD's loser-friendly installation
program to partition my disk exactly the way I wanted. I don't remember the
details, and the disk it was housed on is long gone, so I can't tell you
more. I guess I could have reported these problems to the FreeBSD team, but
by that time I was into VAXen enough to realize that UNIX on IBM PC is not
the way to go, so I didn't bother. That partition laid unused until the
disk was reused for a different purpose. Also if my goal were to get the
problems fixed, rather than merely report them, I would have to subscribe
to one or more FreeBSD mailing lists. I already spend more time on E-mail
than I should, so that's not very appealing to me.
   
> There may be specific issues with certain items of hardware
> support, but you should be able to depend on machine independant
> code without having to extensively test it everytime you add a new
> port, or even processor type.
   
   If you changing some of the key areas like the VM system, you really
have to retest all supported CPU types.
   
> Further on the 'stability' issues - how many stability problems
> does NetBSD/vax have on the _officially supported_ hardware?
   
   There are some problems. What about the clock? What about qe timeouts?
   
> Note the the MV3100 is not yet offically supported - when ragge
> has had a chance to iron out the problems it will be added to the
> supported list.
   
   I don't think that there will be any improvements in the BabyVAX support
unless/until someone takes my patches and tries to port them to NetBSD.
   
> >    - Take NFS from HPBSD 1.x. NFS is absolutely essential for UNIX
> > "clusters". HPBSD 1.x is 4.3BSD vintage, so this should not introduce
> > any impurity problems.
> >
> This would exclude NFS3 which has some real gains.
>
> Hmm, if you go the SunOS route for NFS support that is going to
> introduce 'vfs' layers to support non ffs filesystems. I think
> there may be quite a bit of 'impurity' there, though it would make
> supporting cd9660 and other filesystems much easier.
   
   Does SunOS use the VFS layer? Actually I have just learned that there
are two different NFS implementations in the CSRG/UofU world. 4.3BSD-Reno
introduces Rick Macklem's free NFS version 2 implementation that stays in
Net/2 and 4.4BSD. 4.3BSD-Reno is where VFS is introduced in the BSD world,
so that implementation is understandably VFS-based. OTOH, there is HPBSD
1.x. This baby implements NFS v3.0, so I'm pretty sure that it uses Sun's
genuine NFS code. (This probably means that it requires a license from Sun.
Fortunately, I'm certain that CWRU has one, since we have SunOS sources.)
HPBSD 1.x is 4.3BSD vintage and predates Reno. I think that it somehow
implements NFS without VFS, since I have never seen any references to HPBSD
1.x having VFS. A question to the list: is it feasible at all to support
NFS without VFS?
   
   And what about Ultrix? It supports NFS since at least v3.0. Does it use
VFS? It seems to me that it's not very friendly to non-UFS filesystems.
AFAIK it doesn't support ISO 9660 (implying that Ultrix CD-ROM are written
in UFS). BTW, are MIPSen big-endian or little-endian? I seem to recall that
filesystems are tied to the byte order, and the same Ultrix CD-ROM can
bootstrap a VAX and a MIPS.
   
> Having run a stripped down NetBSD installation on two 20MBs drives
> I can safely say that deciding which programs you do not want can
> also make the difference.
   
   This may do if you are the only user, but if you give the general
populace a heavily stripped-down system they will be upset.
   
   Sincerely,
   Michael Sokolov
   Phone: 440-449-0299
   ARPA Internet SMTP mail: sokolov@alpha.ces.cwru.edu