High Voltage Shareware/high1.zip/high1/DIR21/AUM008.ZIP/FILE5.008

Text File | 1993-09-12 | 30KB | 581 lines

                  NETWORKING AND ASSOCIATED MYSTICISM
           (Culled from messages mainly written by Wes Smith)

        Date: 05-11-93 (09:53)              Number: 1185 of 1204

MA -> well I guess we haven't considered dr dos. I am sure that they will
MA -> soon have a Dr. Dos 7.0 out soon...wonder what that would have? Maybe
MA -> that is the route to go.

Digital Research was purchased by Novell, and DR DOS 7.0 will reflect a
plethora of new options designed to support and enhance NetWare. The most
interesting feature so far is local support for peer-to-peer networking
under an existing Novell network.

Digital planned to ship during the summer, but has recently announced
they'll wait until the Fall.

        Date: 05-12-93 (10:21)              Number: 1186 of 1204

SA -> My question is what type of network is the best in terms of speed,
SA -> reliability, etc for a 2-6 computer network.

At work we have 80 workstations and 3 servers on 2 token ring loops. We use
8-bit 16/4 cards (can be switched between 16-mbb and 4-mbb speeds,
manually) almost exclusively, although there are a handful (ack!) of
microchannel 4mbb-only, and even a few ISA 4mbb-only cards that no one will
admit to having ordered.

TOKEN RING

The Token ring network concept is quite old - the hardware required is
circa WWII and no one's ever felt compelled to bring the technology up to
current standards. Setup is done with a 9-volt battery that tickles
clumsy solenoids.

Both 16 and 8-bit cards are available, in Microchannel and ISA
configurations. Speed is always 16mbb or 4mbb. Even the smallest
installation requires a 8228 MAU (multistation access unit); MAU's come
in 10(8) and 18(16) socket configurations, running roughly $700 and $900
respectively.

    IBM 8228 10(8), 18(16) MAUs  ┌───────────────────────────┐
                                 │   9 0 A B C D E F         │
   ┌───────────────────────────┐ │   ■ ■ ■ ■ ■ ■ ■ ■         │
   │ ╥ r1 1 2 3 4 5 6 7 8 r0 ╥ │ │ ╥ ■ ■ ■ ■ ■ ■ ■ ■ ■ ■   ╥ │
   │ ╨  ■ ■ ■ ■ ■ ■ ■ ■ ■ ■  ╨ │ │ ╨ r1 1 2 3 4 5 6 7 8 r0 ╨ │
   └───────────────────────────┘ └───────────────────────────┘

Each MAU deck requires a line in (r0) and a line out (r1), to form a
physical loop with wire. This is why a 10 port MAU will support 8
workstations, and an 18 port MAU will support 16. When you first install a
MAU, you use an IBM 8288 Setup Aid to initialize the ports. This is a $35
black plastic case for a 9-volt battery, with a red LED and an IBM
4-connector head.

MAUs are mounted in specially-sized (and expensive) MAU racks, rather like
old computer or new stereo components. The r1 socket of each MAU is
connected to the r0 socket of the MAU beneath it. The r1 socket of the
lowest MAU runs to the r0 socket of the highest MAU, forming a complete
physical ring.

A type-4 network MAU rack will also sport an RJ (size is irrelevant; we use
regular phone connectors) multi-plate. On the wall beside each workstation
is an extra phone-like jack. Type-4 wire connects that jack to the back of
a single socket on the RJ multi plate. A special RJ-to-IBM spanner cable
can connect any RJ socket (thus, any workstation) to any MAU socket. This
is important for stream-architecture. With ring network topology, there's
"upstream," and "downstream." Data moves downstream only, and through /
past _each_ workstation. Thus, the MAU socket just upstream of the MAU
socket connected to a server will have the slowest read time and the
fastest write time.  Network guru-types have to struggle with that on
occasion, since noisy workstations (bad wires, poor network hardware)
should be braced between clean workstations, to help wash the signal.
Further, read-only stations (backup) should be immediately downstream of
the server to speed operations. The data-request token takes the long way
around, but the data blocks travel the shortest distance. The overall
chance for frame corruption drops, avoiding the time-wasting need for frame
re-sends.

In our case, we have 2 distinct rings on our MAU rack. I have to juggle
upstream and downstream noise and traffic considerations with basic speed.
Normally I hobble noisy 16mbb stations onto the 4mbb ring - the easiest way
to solve a traffic problem is to cut the throughput by 75%. Speed changes
of this order have to be configured on the workstation side as well, where
a DIP switch selects between 16mbb and 4mbb.

Re: IBM 4-connector heads (AMP 3991). These are gender-less connectors; any
2 heads will connect. Normally, the case breaks apart easily and a clear
plastic, 8-cylinder egg-tray covers 8 metal tubes (each has a graduated
slice in the side, to cut and hold an un-stripped wire).

   ┌────────┐  ┌┐┌───────────────┬┐      ┌───┬┬───────────────┐┌┐
     ▄ ■─ ▄ │  │├┤ ┌───┐   ││ ┌┐  │┌┬┐││     ┌───┐ ├┤│
   │ ▀ ■─ ▀    │││ │   │      ├┴─┴┤  └┘││││      │   │ │││
   │ ▄ ■─ ▄    │││ │   │      ├┬─┬┤  ┌┐││││      │   │ │││
     ▀ ■─ ▀ │  │├┤ └───┘   ││ └┘  │└┴┘││     └───┘ ├┤│
   └────────┘  └┘└───────────────┴┘      └───┴┴───────────────┘└┘

IBM recommends (IBM) type-2 cabling. It's about as thick around as your
baby finger, 99% insulation by weight. We use type-4 phone cable. It works
just as well, although IBM refuses to support it.

Token Ring adapter cards have 9-pin DIN connectors on the back. Another
expensive and should-have-been-unnecessary device is required to connect the
TRAC to the wall-jack: a Media Filter Kit. It's basically a head-type
changer, with a few components inside to justify its existence. My opinion
is that signal processing should be done inside the card itself. I suppose
this provides an added compatibility, but I'd rather save the money.

Token ring is robust, but slower than it could be. No matter what size your
network is, only 1 station / server can be non-passive at any given
instant. The applies a practical limit of 96 stations (and 12 IBM 8228
MAUs) to any given ring. Data is transferred with Request, Send, Verify
through layers of protocol software. The ring topology means that all data
passes through each TRAC. This raises a considerable security problem, and
it is indeed possible to monitor the data flow, picking out passwords and
interesting memos, at will.

No single piece of TR hardware can be more than 150' from the next piece,
and the maximum distance on the MAU patches is limited to 400'.

Token Ring is too expensive (16/4 cards run $500+), and ultimately limited,
to be taken seriously. IBM still pushes it, for the usual IBM reasons:

   1. It's expensive ($$$ for IBM).
   2. Not many competitors, since the whole thing is too ugly.
   3. The parts aren't upgradeable (changes mean $$$ for IBM).
   4. The parts aren't compatible with other nets (can't change from IBM).

Token Ring can still be found in many, many companies and it's a very good
skill to have. People who need Arcnet or Ethernet techs just ask for
network people. Token Ring installations specifically ask for Token Ring
Network Gurus when they hire.

        Date: 05-12-93 (10:17)              Number: 1188 of 1204

SA -> My question is what type of network is the best in terms of speed,
SA -> reliability, etc for a 2-6 computer network.

The FlatEarth runs on a 5 station, peer-to-peer Ethernetwork, wired with
thin Ethernet (Cheapernet) cable. All of the network hardware is on the
actual cards - Ethernet doesn't require a hub or repeaters for a basic
configuration.

ETHERNET

Ethernet networks are relatively cheap, relatively fast, and quite common.
The only real differences between various Ethernetworks lies almost
exclusively in the wiring. There's a cheap way to do it, and a grossly
expensive way to do it. You get to use the cheap way unless:

   1. Large distances need to be covered.
   2. Your network contains a gateway to a non-PC network.

Cheapernet, 58Ω co-ax wire, runs from machine to machine in a straight line
(string topology is used in Mac networks too, incidentally). Generally,
it's easier to stay on top of your wire budget (both financially, and
geographically) than with a ring topology.

Supposedly, you can only bring a workstation onto the wire through the use
of a BNC T-connector. I'm not satisfied that a star-node connector wouldn't
work, although I haven't had enough connectors to test the idea. Fritz
swears it won't work, and he's probably right.

The second wiring method (the expensive one, Thick-Ethernet) requires a
transceiver box for each workstation and attaches to the DIX connector
socket on the back of the Ethernet adapter card. The topology is the same,
but you need to use N-series connector plugs and terminators instead of BNC
hardware. The transceivers alone will run you more than a cheapernetwork.

With Thin Ethernet, you're limited to 30 stations per trunk, with a maximum
distance of 600' between workstations. It's possible to link multiple
trunks together, but you need repeaters and bridges. The maximum length for
Cheapernet is 3,000'.
Thick Ethernet can go 1,600' between workstations, with an overall network
length of 8,200'. Again, multiple trunks can be linked with special
hardware.

Interestingly, it's possible to mix Ethernet cabling types. This requires
more repeaters and converters than you if you picked one over the other,
but demonstrates that the original Ethernet designers were focused on
compatibility. That's very important, especially if you're using gateways
and bridges to link in non-PC platforms.

Ethernet cards can be quite cheap, and there's a good selection to choose
from. I've purchased 4 CNet 100e500e 8-bit 2mbb adapters; Fritz "found," a
3com 16-bit adapter that was willing to work at 2mbb, and we brought it
onto the network as the 5th node. I pay $100 through a vendor / friend, but
I expect that they must be available for much less, somewhere.

Ethernet cards of different speeds can't talk to one another, but they
can share the same wire. It's possible to string a 2mbb and a 16mbb
Ethernetwork together on the same length of co-ax. They won't even see each
other, and there won't be any interference.

        Date: 05-12-93 (10:17)              Number: 1189 of 1204

SA -> My question is what type of network is the best in terms of speed,
SA -> reliability, etc for a 2-6 computer network.

ARCNET, etc

I have very little experience outside of Token Ring and Ethernet. The first
network I ever used however, was Arcnet. The controller relayed various
woes, most being directly related to the lack of a proper industry
standard, and the pathetic throughput.

My understanding is that it's difficult to run two Arcnet cards from
different manufacturers, in the same network. Further, I _think_ that
Arcnet requires (at least) a passive hub; you can't just link the cards
together the way you can with Ethernet.

I have no knowledge of Arcnet costs.

IBM used to market the "IBM PC Network Adapter," the "IBM PC Cluster
Adapter," and the "IBM Personal Computer Data Acquisition and Control
Adapter," each of which has a network application. The IBM PC Network
Adapter can still be found in service, as well as a derivative, the "IBM PC
Network Baseband Adapter." I couldn't even begin to comment on these,
except to point out that no one but IBM supports them. That implies bulky
standards, high costs, and uninspired technology.

        Date: 05-12-93 (10:22)              Number: 1190 of 1204

SA -> My question is what type of network is the best in terms of speed,
SA -> reliability, etc for a 2-6 computer network.

GENERAL

All PC network software operates in much the same way. The cards rely on an
IRQ and a Base I/O address, or an IRQ and a system address. The first TSR
loaded is a driver for the card, making it available to DOS. The second TSR
(or more likely, group of TSRs) manages what are, effectively, the virtual
resources available through the network.

Under Novell, the tech uses a utility called NETGEN to build a NETX.COM
program that is then distributed to each workstation. It combines various
revisions of several other drivers and extensions, linking them together
into a single file. The NETX.COM program from one NetWare installation will
very probably not work on another NetWare network.

Under LANtastic, the network environment is set up by loading AILANBIO (a
smaller, smarter derivative of Novell's NETBIOS), REDIR to extend DOS to
handle virtual resources, and SERVER (if applicable) to make local resources
available to other workstations. DOS's SHARE.EXE provides file and record
locking.

The NetWare scheme is very typical of Client-Server networks; the LANtastic
scheme is very typical of Peer-to-Peer networks. Both C/S and P2P network
software abounds today. The accepted market leaders are Novell and
Microsoft, of course, although LANtastic has been rated as the best
functionally, and technically, for several years running, now.

Client Server networks are appropriate for large installations, where
you've got a high-end machine with huge resources available to dedicate to
running things. You can't do much on a main server under Novell.

Peer to Peer networks are handy when you're dealing with a small group of
workstations, or where the basic draw on the network is slow.

A C/S installation can rescue a fleet of older ATs by providing each with
next-to unlimited storage, and group resources. A P2P installation can
save your budget by making every single piece of hardware you have,
available to every computer you can link into the net.

OFFICE

Our office uses a Novell 3.11 network with 3 servers. The largest / fastest
(Compaq 486/66, 100 user license) supports everyone - it holds all the
basic applications and user data. A second server (IBM PS/2 m80, 50 user
license) runs AccPac, and nothing else. The third (IBM PS/2 m80, 20 user
license) is the exclusive domain of Tech - I use it to prototype software,
network variations, etc.

Each workstation runs IPX and NETX to bring it into the network. These
consume 60K of RAM, and can be loaded high. Most workstations are also
running LANtastic for Novell, a clever piggy back system that installs by
using NetWare's own ODI as the network adapter instead of requiring an
actual card. This consumes another 50K, plus 50K for NETBIOS. Artisoft's
own NETBIOS derivative, AILANBIO is a fraction of that size, but can't be
used in Novell's environment. That last 50K is the weakest part of the
chain, since it can rarely be loaded high (not many 50K wide upper memory
segments available).

Our resulting environment has 2gig available across the 3 main NetWare
servers, and allows Workgrouping with the 1gig of local HD space on the
workstations, which had gone unused after the original Novell installation.
The main logon script (a batch file, of course) is an elaborate affair that
determines the LANtastic Workgroup connections required for any user
logging into any Novell workstation. Connection definitions are effected
with zero byte files in directories on the main Novell volume. Users are
set to run their applications from the NetWare volumes, and save their work
on their LANtastic links (local HDs on computers in their own department).
Deferring the vast bulk of (generally) useless user files to the slower
local HDs and using the expensive main servers exclusively for fast
application loading, met with approval from all quarters.

HOME

The FlatEarth is running 5 systems under LANtastic 5.0, with Ethernet cards
and Cheapernet cabling. Ethernet is far, far simpler to use than Token
Ring: there's no hub, no fancy wiring or extra hardware required. Artisoft
sells, besides the LANtastic for Novell, a slotless version that works with
serial and parallel cables, as well as modems. Beyond that, they sell
several specific-to-Artisoft-Ethernet-cards versions, and their /AE
package, which will run on almost any manufacturer's network cards (that's
what TFE uses). Further, there's a version for Windows, that I've not yet
looked at. I expect it's Artisoft's answer to Workgroup for Windows, which
is Microsoft's answer to NetWare for Windows, which is Novell's answer to
Windows, which is Microsoft's answer to Macintosh, which was just a really
cool idea Wozniak had one day while cleaning out his garage.

        Date: 05-12-93 (10:22)              Number: 1190 of 1205

SA -> My question is what type of network is the best in terms of speed,
SA -> reliability, etc for a 2-6 computer network.

GENERAL

All PC network software operates in much the same way. The cards rely on an
IRQ and a Base I/O address, or an IRQ and a system address. The first TSR
loaded is a driver for the card, making it available to DOS. The second TSR
(or more likely, group of TSRs) manages what are, effectively, the virtual
resources available through the network.

Under Novell, the tech uses a utility called NETGEN to build a NETX.COM
program that is then distributed to each workstation. It combines various
revisions of several other drivers and extensions, linking them together
into a single file. The NETX.COM program from one NetWare installation will
very probably not work on another NetWare network.

        Date: 05-12-93 (10:20)              Number: 1191 of 1205

SA -> My question is what type of network is the best in terms of speed,
SA -> reliability, etc for a 2-6 computer network.

STRATEGY

On the most basic level, networks are a way of assigning a local drive
letter to a virtual resource. NetWare will allow you to use drive letters
_above_ whatever your CONFIG.SYS's LASTDRIVE is set to, while LANtastic
will use letters _below_ the same limit. The trick is not to let the
network aspect overwhelm you. Too many new-to-Novell techs fire off drive
letters for application and data paths without reservation, with the chief
result of confusing the network's users and administrators. Simply treat
each resource as a single drive letter (avoid having multiple letters
pointing to the same volume) and use security to handle limit access. In
addition to lifting the veil of confusion from network mapping, this also
has the advantage of using the smallest number of drive letters, which
you'll find to be in high demand on networks with multiple resources.

You need a strategy to set up your maps. Physical devices start at a: and
work their way up, so you should start your virtuals at z: and work your
way down. When they meet, it's time to switch to UNIX. For a specific-
application environment, like a network or a software server, use virtual-
virtual maps. PCBoard should be installed on its own drive letter, since
the setup requires explicit pathnames for scores of data files. You can
move the entire virtual mapping to another virtual resource without having
to change all the explicit pathnames. Maps from TFE's main server, TYCHO:

Real (not-network):

   a: local floppy
   b: local floppy (remaps to a:, no second floppy)
   c: local IDE
   d: local IDE
   e: local SCSI, driver map
   f: local CDROM, driver map
   g: slave floppy ┐
   h: slave floppy ├─ Effected through Brooklyn Bridge and a null modem,
   i: slave RLL    │  connected to a not-networked XT (old NEWTON) with 2
   j: slave RLL    ┘  HDs. These slave-to-Tycho resources are available
                      to LANtastic for resource mapping as thought they
                were normal local drives.

Virtual (network):

   m: TYCHO's d: IDE
   n: TYCHO's e: SCSI
   o: virtual "virtual", based on TYCHO's d: IDE
   p: NEWTON's c: RLL, slave to Tycho as Tycho's i: virtual
   q: NEWTON's d: RLL, slave to Tycho as Tycho's j: virtual
   r: FERMI's c: RLL
   s: FERMI's d: RLL
   t: BOOLE's c: MFM
   u: BOOLE's d: MFM
   v: EUCLID's c: MFM
   w: EUCLID's d: MFM
   x: TYCHO's c: IDE
   y: virtual "pcboard", based on TYCHO's d: IDE
   z: virtual "login", based on TYCHO's d: IDE

These started in order, but as the network changed, server drive mappings
became separated. It's important to divorce yourself from the idea of the
necessity to keep things perfectly linear - it doesn't help the network,
and you'll give yourself an endless amount of worthless maintenance,
remapping everything and changing the defaults in all your programs.

Notice the 3 virtual drives. "Login," is simply a subdirectory on TYCHO
that each workstation maps to when it first links to the network. It
contains a large common batch file; by keeping a minimum of instructions in
each station's AUTOEXEC.BAT and having them all run SURROUND.BTM (on the
network) I only have to make startup changes once to affect all the machines.

"PCBoard," contains the BBS and doors. All paths in PCB's setup point to
the y: drive. When I had to move the entire tree to a different HD, I
simply re-wrote 2 lines in the common batch, SURROUND.BTM to say that Y:
was on TYCHO's IDE instead of the SCSI - no changes were made to PCBoard at
all.  Virtual-virtual paths require less effort to change than simple
virtual paths. That's important.

The "Virtual," virtual-virtual is used for installing applications. UnZIP
the various disks to their own subdirectory on the same volume (but in the
non-virtual-virtual path), and swap the files into, and out of "Virtual,"
while running the install from another machine. This only fails if the
install program reads disk labels.

Get out of the habit of thinking of your PC as having c: and d: drives.
Think of them only as their virtual map names, perhaps x: and y:. That will
prevent confusion - they're x: and y: _everywhere_ but c: and d: only
locally. Each resource should be the same letter on each machine, without
exception. Accomplish *ALL* mappings with a single batch, that all
workstations run at login time.

Batches load and run faster over the network than programs. Most operations
can be handled by batches, especially if you give the system a boost by
switching to 4DOS BTM's, which runs scripts from RAM, rather than disk.
Batches can check environment variables and make decisions easily; expand
your environment size and use variables freely. Make the server's name,
Fido address, modem speed and port, etc ... available as variable, and have
your batches respond intelligently. At home, the batch that runs the BBS
will exit with an error unless the %SERVER% variable is FERMI or BOOLE.

Security is a science unto itself, and is probably not going to be
important to you with a simple home network. For a LANtastic network, give
the default user account, "*," full Supervisor access and leave it at that.
If you ever have to manage a large commercial network, it can give you
endless nightmares, but we can discuss that at a later date. ;+)

        Date: 05-12-93 (10:32)              Number: 1193 of 1205

PB -> SCSI CD-ROM? Scsi is simply an interface type, I am not aware of any
PB -> CD-ROM's that DON'T work from SCSI.

Several manufacturers, most notably, Mistumi, ship their CDROMs with a
proprietary controller card. My experience has been that these prop cards
are unlabeled, considerably hobbled, SCSI cards. Your statement stands,
but not without modification, since prop-SCSI might as well be it's own
standard - it's incompatible with anything else.

        Date: 05-12-93 (10:33)              Number: 1194 of 1205

FTC -> ... or wait for D.R.DOS 7 which
FTC -> should be the replacement for netware lite.

Ugh! <clunk> Not quite; let me explain this one more time. DR DOS 7.0 will
enhance an existing NetWare Client Server installation by providing Peer to
Peer access, likely through the Novell ODI.

ie: To get Peer to Peer, you have to already have Client Server.

        Date: 05-12-93 (10:36)              Number: 1195 of 1205

SA -> Hi...umm...is there any good books on Networking in general (how to
SA -> setup a network, what software works with which cards, cabling, etc
SA -> etc) that anyone could recommend.

Ranger has recommended Microsoft's Workgroup for Windows Technical Manual.
I saw it for a second, and it looked OK. Supposedly it's only $30, too.

I don't think I've seen a general (and still useful) network primer. In
fact, I don't think I've seen many intelligent books on network hardware
and strategy at all. If you find one, I'd love to hear about it.

        Date: 05-12-93 (10:38)              Number: 1196 of 1205

TA -> NDOS is a COMMAND.COM replacement. It is smaller than MS and IBM DOS
TA -> COMMAND.COM, and is more user friendly.

BALLS! NDOS (actually, 4DOS) is a COMMAND.COM replacement, but it was
designed to add programming and operations power to DOS, *NOT* to make it
easier to use, or user friendly. It adheres to Microsoft's OS
specifications, which DOS _doesn't_, adds a wealth of additional commands
(including some very intelligent batch functions), and offers up access to
hardware and environment conditions through system-maintained variables.

        Date: 05-12-93 (20:53)              Number: 1197 of 1205 (Echo)

»» Weren't you advocating NT?  How much have you used it?
»»
»» I've played around, and read up enough on OS/2 to get a feel for it.  All
»» I'm saying, if is you expect NT to be better, you'll be disappointed.
»» Extremely hardware hungry, same Windows interface, and hasn't gone through
»» the bug revisions as much as OS/2 of course- heck, of course- it's not even
»» out yet.

Well I got the beta a while ago, and now I killed it.  So you win.  It
couldn't do dick shit with some programs already, and I heard it soon wouldn't
support any IBM stuff at all, so I guess I'll have to look deep into the New
version of OS/2..

        Date: 05-12-93 (20:55)              Number: 1198 of 1205 (Echo)

»» It is too bad that many software companies don't get along, especially ones
»» that hold has much of the main markets out there as Microsoft and IBM.
»» Windows /NT is rumoured to be able to support OS/2, Windows 3.0, 3.1 and
»» Windows /NT applications, as well as DOS applications.  However, I wonder
»» if the competition btwn IBM and MS will have an effect on how WELL IBM
»» applications (DOS/OS/2) are supported in MS's Windows /NT.

I don't think so though, since MicroSoft and IBM are splitting up and IBM
ain't getting the source from MicroSoft and vice versa, there's going to be a
major conflict between winning people to buy any products from that specific
company. (Hey, this might be the Terminator ]I[ we're waiting for eh? :))

        Date: 05-12-93 (20:57)              Number: 1199 of 1205 (Echo)

»» The message is:  Read fault error reading drive c.  Abort Retry Ignore
»» Fail?  I haven't had any other problems with any other programs, no viruses
»» scanned for (using MacAfee's 104) were found, and no diagnostic programs
»» report any errors.  It is the strangest thing I have ever encountered...

Read fault error?  Do you have a bad sector in your C? Have you tried NDD?
Or maybe mem couldn't understand your config.  I dunno, try getting another
copy of MEM, see if it still prompts for that message.

        Date: 05-13-93 (10:23)              Number: 1200 of 1205 (Echo)

»» I don't think so though, since MicroSoft and IBM are splitting up and IBM
»» ain't getting the source from MicroSoft and vice versa, there's going to be
»» a

    There will always be competition between IBM and Microsoft, but I think
the selling points in any operating system are compatibility.  If IBM and
Microsoft don't make things compatible, than they may lose business.

        Date: 05-13-93 (10:25)              Number: 1201 of 1205 (Echo)

»» Read fault error?  Do you have a bad sector in your C?  Have you tried NDD?
»» Or maybe mem couldn't understand your config.  I dunno, try getting another
»» copy of MEM, see if it still prompts for that message.

   It is the strangest thing, I have run NDD, and Stacker's check, and even
CHKDSK, but no disk errors were found.  I don't see why MEM would have a
problem with the config.sys file, as it checks the memory addresses, and
gathers information from the system resources, rather than the config.sys
file. I have replaced MEM a couple of times from the original, but it still
screwsup.  Very odd.

        Date: 05-13-93 (10:28)              Number: 1202 of 1205 (Echo)

»» BALLS! NDOS (actually, 4DOS) is a COMMAND.COM replacement, but it was
»» designed to add programming and operations power to DOS, *NOT* to make it

   It does add some useful tools from what I have seen.  I don't have it
installed, but I was playing around with it.  The only real advantage that I
saw was the user friendliness compared to the MS DOS interpreter.  It seems
simpler to use for a computer illiterate.  However, its utilities that support
batch files are supported through Norton's batch enhancer program, so I think
I'll stick with the regular real mcoy.  What caught your fancy with NDOS/4DOS?

        Date: 05-13-93 (13:14)              Number: 1203 of 1205 (Echo)

It's worth it if you don't use Desqview. I use emm386.exe and with memmaker,
it gave me more memory. It doesn't conflict like qemm does, but if you are a
desqview user, stick with qemm.

        Date: 05-13-93 (17:04)              Number: 1205 of 1205

-> what TFE uses). Further, there's a version for Windows, that I've not
-> yet looked at. I expect it's Artisoft's answer to Workgroup for
-> Windows, which is Microsoft's answer to NetWare for Windows, which is
-> Novell's answer to Windows, which is Microsoft's answer to Macintosh,
-> which was just a really cool idea Wozniak had one day while cleaning
-> out his garage.

   Most of your previous messages I've just captured without comment,
but I must say that this little gem was so funny I fell off my chair.
If I broke any bones I think I'll sue for reckless misuse of a joke. :)