Network OS's (Was:Off Topic - Novell Contractor Needed)

[Jonathan Hutchins, Rune Webmaster]

----- Original Message -----
From: "Jeremy Fowler" <jeremy at microlink.net>

> Ok, I understand you may have had prior bad experiences with Novell, and I
can
> respect your opinion, however so much that I disagree with it. I know
there is
> nothing I can say to sway your opinion, so we must agree to disagree. One
> question though, why NT? Why not Linux, or FreeBSD, or OpenBSD? I don't
believe
> NT is a viable solution to a production network anymore than you think
Netware
> is.

OK, long reply, based on about eight years of actual field experience with
these systems:

NT's pretty good for small to medium sized networks, and because it can use
straight TCP/IP it becomes a non-issue in larger ones.  I'm a local admin
for a system where we run NT all over the country - nearly 5,000 systems in
my domain alone, which is one of two primary domains and among 184 domains
in all.  We converted from Novell to NT in late 98 - early 99, and saw a
significant rise in productivity and reliability.  Network capacity
increased significantly when we eliminated IPX traffic.  I had a few systems
that lingered on Novel for a while, so I was able to see side-by-side
performance differences.

NT is used for our file and print servers, and some light database work.
Our heavy servers are mostly IBM systems running TSO or AIX; and we have a
number of "conventional" Unix systems (meaning I don't know what hardware or
OS flavor they are, I just link people to them).

We've found NT to be extremely stable - the servers here have been running
continuously since late 1998 without a crash, and the majority of
workstations simply don't have problems.  We had one rash of non-Windows
software crashes which we were able to solve by increasing the RAM to 128M.

On my home system I run Windows95 workstations and both an NT server and a
(RedHat) Linux server/firewall.  I chose 95 because it's less resource
intensive than NT and I don't need NT's security features, and I don't see
any advantages in the later less stable Windows releases.

I've also found Windows 95 to be extremely stable and reliable if installed
on decent hardware, and if one is conservative about what software is thrown
at it.  I used to run a FIDOnet BBS on Quarterdeck's DesQview, and moving to
Windows95 was a real boost in performance and reliability - the system ran
continuously for 18 months at one point without any crashes or problems (and
would have run longer if I didn't so enjoy messing with it).

My NT server is my primary server because I've found file/directory
permissions were easier to manage than the Linux/Samba permissions.  This is
more likely a shortfall in my own knowledge than and inherent problem with
the system, but there it is, it was easier to get the 95/NT permissions
right.

In the long run I think we'll probably see Linux appear on office desktops
within two to three years.  Microsoft's trend toward "co-branding" to lock
in business partners and force users to buy from them; their increasing
code- and feature- bloat, and the dumbing down of their systems is
eventually going to force corporations that aren't development partners to
look elsewhere.  With distributions like Mandrake 7.2, we're getting close
to a desktop that could realistically be rolled out to a staff that was
"trained" on Microsoft products without much productivity loss or changeover
problems.

But Novel?  Novel always had great tech, and they always benchmarked really
well, and they always failed in the field.  They're an interesting
historical footnote - the first real, practical PC Networking system.  But
they've never really come into the TCP/IP generation.  They were slow to
adopt industry standards over their own proprietary methods, and as badly as
Microsoft has done at mangling standards, they've done better at adopting
them than Novel did.  Novel never managed to integrate with Windows without
hanging and crashing.  Every office where we installed NT beside an existing
Novel network got rid of Novel in less than a year.

Once PC's started to adopt Windows (3.1), Novel started to slip.  The
drivers were difficult to load in DOS memory, and would often cause
problems, hang, or fail if loaded to high memory.  Trying to add Novel's
implementation of TCP/IP was a real nightmare, but it could be made to work.
As I recall it provided a pretty effective firewall, but that was before
anybody was really worried about that.  When Windows95 came along, Novel
seemed to take the attitude that it was just a fad, serious work was done on
text-based interfaces and this GUI stuff was for game players.  They caught
their mistake too late, and they never managed to catch up.

Networking today means TCP/IP over ethernet.  It means desktops that are
Microsoft because that's what everybody knows how to use.  It means servers
that provide Windows interfaces that work with Windows API's and who's
operating systems are invisible to the end user.  NT, xNIX, IBM and even DEC
VMS servers do just that, and do it well.  The choice among them is based on
what software the server needs to support, and what the scale will be.  NT
has a definite place in the mix, and integrates reasonably well.  Linux,
BSD, and other Unix flavors have huge advantages in scalability and
flexibility.

In the long run, anything that isn't xNIX will probably look like it.
Desktops will probably become more like appliances, with the exception of
power users who are already adopting Linux.  Microsoft won't go away, but
it's already lost the war, and it will loose it's market dominance sooner
than most people expect.