Old farts reminiscing about computers

[scaesare]

Oh gawd... Don't remind me of WFW, Win3.11, and NT "networking". So many jumpers on cards, crazy config files in the correct order, netmask and gateway settings that didn't follow what unix was doing. It was a mess to set up a heterogenous network with unix and windows PC trying to co-exist. And then there was "token ring". Barf.

I think Microsoft was playing "catch up" with the whole ethernet thing back then. And IBM was pushing to make Token Ring the standard, and have Ethernet go away.

I worked in a building that was wired for "USOC" standard RJ45, so the new ethernet cables required an adapter to plug into the building wiring.
Standard EIA /TIA 568 A EIA /TIA 568 B, USOC, RJ45 Pinout Wiring Diagram

(At least that was better than installing "vampire taps" into fat coax Ethernet.)

Certainly true during the Win3.1 days, where the network stack was basically a real-mode set of drivers you had to cram in to memory. And ISA cards and jumpers were indeed a challenge... Although standardizing our 3C509's on IRQ 10 and making sure any internal modems didn't share IRQ's, but were rather on IRQ5 simplified a lot.

And the emerging HW standards wars made things interesting too.... 100Base-T or HP's 100Base-VG "AnyLan"? IPX/SPx pr IP? Microchannel or PCI? VESA LocalBus?

But my point was that NT and WfW made that process significantly easier. Plant your net card on an open IRQ, and load the WfW drivers for the card... no more futzing with HIMEM statements, etc... NT even more so with it's 32 bit model. PCI cards were coming in to play then, and Plug N Play took a big step forward.

Interoperability actually took a huge step forward at that point. Our mid-sized company eventually had:

-NT Server for domain/file/print services using the native IP stack for client connectivity
-NT talking to existing printers using the native DLC stack
-WfW accessing file/print services using native IP stack and Lanman client
-UNIX hosting SAMBA shares for WfW client machines without having to load separate client software
-Mac clients accessing NT shares via NT's Services for AppleTalk
-NT Server mounting AppleTalk printers to share to clients using it's native Mac services and a DayStar digital AppleTalk card (drivers in the box for NT)
-NT & SNA server providing gateway services for WfW clients w/ 3270 terminal software to talk to IBM midranges (System/36 and AS/400)
-NT providing DHCP/WINS for the whole network
-NT providing gateway access to the development Novell network to WfW clients via Services for Netware and it's native IPX stack (no client software needed on workstations)
-NT providing RAS services to global remote offices via WfW built in Dial-up-networking profiles connecting to IBM's X.25 packet network
-NT providing internal company mail via Microsoft mail internally, and then using Post.Office for SMTP internet connectivity


The vast majority of the above was made possible with what NT and WfW brought to the table. Very little third party software was needed. Things were vastly simplified, and the model was a single standardized client & stack, with services and gateways moved to a server where possible. Administration at sigle points on a server rather than at 100's of points on clients.

The company (~170ish employees) took a gamble on me as a 22 yr old kid and let me implement this for them.

Prior to that they had a hodge-podge of separate networks and islands of connectivity: A UNIX cluster for engineers with some workstations and some PC's and an atrocious Sun NFS client. Som PC's had 3270 cards attaching to TwinAx . AppleTalk cards in others to talk to LaserWriters. Real 3270 terminals scattered about. A Novell network island for developers. Mac's strung together with Farallon PhoneNet connectors to talk to AppleTalk printers. PC's that needed to connect to multiple services were a mess.

By the time I was done, there was a standardized client PC configuration that could talk to everything on a single network. And for significantly less $$$ than the consultants who wanted to install a brand new Novell network were gonna charge.

Probably some of the most fun and greatest professional development time in my life. I was largely left to figure it out on my own and loved every minute of it.

 

[TEG]

HIMEM. Ugh. Ramdisk.sys Ugh.
“Oh my network stopped working... all I did was install a mouse card!”

Too bad we didn’t have someone like you to write custom drivers for us.

 

[scaesare]

Don't get me wrong... hardware on the PC with DOS was ugly. Getting it to work with Win3.1 bolted on top was uglier.

But my point was addressing the 'Don't remind me of WFW, Win3.11, and NT "networking"' statement. Networking was coming, there was no stopping it... and that meant more hardware to install/configure, and more demands on the OS and memory for drivers and client software.

WfW and NT were the first steps in the PC world at alleviating much of that hassle. Lumping them in with Win 3.1(1) isn't accurate, IMO.

 

[HankLloydRight]

What was that program that would try every combination of loading DOS drivers into lower or extended memory to get the most optimal fit (and still work)?

Man we had fun with that program.

 

[mongo]

What was that program that would try every combination of loading DOS drivers into lower or extended memory to get the most optimal fit (and still work)?


Man we had fun with that program.

I used Wolfenstein to check XMS/ EMS

Bought a 57.6k modem freshman year. Then the next year the dorms had ethernet (upgrade from serial line drivers). 3c509 driver with menu based autoexec and config.sys for tuning based on program you were going to run. (Mostly x-wing). First one on the hall with cd-rom drive and SoundBlaster card, came with Kings Quest 6. Thought I broke the computer installing the drive. Turned out I'd swaped the ps2 keyboard and mouse...

At work study job at U of M we had a PC with 64 MB of RAM. For grins, I made a ram drive and copied the Win 3.1 directory to it. Windows was running before the monitor finished changing resolutions.

Remember Doublespace for DOS 6? More like all your space since it would corrupt if you ran chkdsk on it (or looked at it sideways), and all your data went poof.

Win for workgroups was handy for instant messaging... and playing hearts.

/ramble off

 

[HankLloydRight]

Remember Doublespace for DOS 6? More like all your space since it would corrupt if you ran chkdsk on it (or looked at it sideways), and all your data went poof.

I once had someone ask me if the could run Doublespace twice to get even more disk space savings. Yes, it's true.

 

[wdolson]

Don't get me wrong... hardware on the PC with DOS was ugly. Getting it to work with Win3.1 bolted on top was uglier.

But my point was addressing the 'Don't remind me of WFW, Win3.11, and NT "networking"' statement. Networking was coming, there was no stopping it... and that meant more hardware to install/configure, and more demands on the OS and memory for drivers and client software.

WfW and NT were the first steps in the PC world at alleviating much of that hassle. Lumping them in with Win 3.1(1) isn't accurate, IMO.

One of the biggest reasons Win95 was such a hit was Microsoft made networking a lot easier to set up and manage. It didn't hurt the interface looked better. They have been chasing that big a hit ever since, but there just isn't as big a leap left in the OS world.

Back when I started out in the work world in the late 80s, I worked in a lab at Boeing that started doing digital simulation back in the early 60s to verify all the electronics on the 727. The lab was in the same place where it had originally been built, but we moved around 1993 and had to sort and pack up almost 30 years of stuff. Some of the systems still in use had incomplete documentation that made them tricky to move.

When cleaning out one area I came across a catalog from the mid-70s for parts for the simulation computers they were using then (some were still in operation in the lab). I remember a spare memory board with, I think, about 4K of RAM was $7000. It was worth a laugh then, but few people would believe you today.

I recall we found some old ferrite core memory in storage somewhere. That place was like an Archeological dig.

They moved us from one of the buildings built in Renton for the B-29 program to a brand new building across from Boeing field. The lab had gone from around 150 to 1500 people for the 777 program and we were stumbling over one another, crammed together. The new building was massive and there were still empty labs when I left in late 94. The lab I worked in went from about 200 sf to about 4000.

For the 777 testing I had to write the device driver for an ARINC 429 simulation board that had a 68040 on it. It was one of the first pieces of hardware using that processor. One of our engineers was the world's best troubleshooter. Tell him this one light comes on every Thursday at 2 PM for no reason, but no other time and he would find it. He found a slew of errors in the 68040 microcode and was on the phone with Motorolla a couple of times a week.

We had banks of 19" racks with card chassis in them interfacing the box for the aircraft under test with the simulation computer. Another problem we had was after running continuously for a few days, various chassis would have random faults and cards would start crashing. Our troubleshooter ran that to ground, and it was literally a ground problem.

Switching power supplies have a periodic small current spike that gets induced onto the incoming line. It can cause noise problems in very sensitive environments, but most of the time people ignore it. Another thing switching power supplies do that we didn't realize until we got a lot of them is if left on long enough, they all synchronize with one another, so the spikes happen at the same time.

In many systems we were seeing spikes of a couple of hundred amps, and I think the largest system had around 1500 A spike. It was amazing the systems kept working at all with those spikes.

They went on a systematic mission to put filters on the power lines into every chassis and the noise dropped down to acceptable levels.

I like working in small companies better than a behemoth like Boeing, but the work could be fun.

 

[mrElbe]

On single sided floppy disks we cut a notch on the opposite side to use them as double sided disks.
Also paid $ 500.00 for a Creative Labs single speed CD ROM drive.

 

[mongo]

On single sided floppy disks we cut a notch on the opposite side to use them as double sided disks.
Also paid $ 500.00 for a Creative Labs single speed CD ROM drive.

Oh the memories! I used a drill press to convert 3.5" disks to high density.

 

[scaesare]

We used a hole punch

 

[scaesare]

I was using Windows 3.11 (Workgroup Edition), then switched to NT 4.0 as soon as it came out (my first PC had "just" the minimum RAM required to run it - 12 MB). Multi threading and multitasking, true 32-bit, SCSI support (had an excellent Pioneer CD-ROM and Philips CD-RW) and networking won me over 95. Even Diablo ran better on NT than on 95

I always wanted to get SMP but could not justify it. What a real pain to run 3D renders on a single CPU.

Same... I was an NT convert while Win95 was the mass market OS that was being adopted.

I cheated and ran the Win95 preview shell for a while.... altho it was buggy.

 

[scaesare]

One of the biggest reasons Win95 was such a hit was Microsoft made networking a lot easier to set up and manage. It didn't hurt the interface looked better.

Win95 certainly took another step in simplifying getting connected. But the real impetus for Win95 was it was the first mass-market OS marketed to users to implement a full 32-bit API (despite continued 16-bit underpinnings).

That was by far the biggest architectural differentiator. It promised no more DOS config file tweaking, no conventional memory issues, threading, preemptive multitasking, Plug N Play, mobile support w/ power mgm't, enhanced graphics subsystems, and a shiny new GUI. It delivered on most of that anyway.

 

[swaltner]

We used a hole punch

You Neanderthal! You should have used something like this to ensure proper placement of that hole.

 

[mongo]

Win95 certainly took another step in simplifying getting connected. But the real impetus for Win95 was it was the first mass-market OS marketed to users to implement a full 32-bit API (despite continued 16-bit underpinnings).

That was by far the biggest architectural differentiator. It promised no more DOS config file tweaking, no conventional memory issues, threading, preemptive multitasking, Plug N Play, mobile support w/ power mgm't, enhanced graphics subsystems, and a shiny new GUI. It delivered on most of that anyway.

For tales of the contortions that engineers at Microsoft went through when creating new versions of Windows (along with other such tales), check out Raymond Chen's blog at The Old New Thing

 

[scaesare]

For tales of the contortions that engineers at Microsoft went through when creating new versions of Windows (along with other such tales), check out Raymond Chen's blog at The Old New Thing

Yeah, Chen's blog has some great stuff in it.

Custer's "Inside Windows NT" is pretty interesting reading as well... she had access to Cutler, Perazzoli, Lucovsky, and a bunch of the ex-DEC boys that basically re-implemented VMS as NT. It's much more about the product than the people, though.

Pascal's "Showstopper" is more about the project itself. Has a bit of the "Soul of a New Machine" feel to it.

 

[CuriousG]

I once had someone ask me if the could run Doublespace twice to get even more disk space savings. Yes, it's true.

Was that the same person that would use a CD tray as a cupholder?

 

[scaesare]

You Neanderthal! You should have used something like this to ensure proper placement of that hole.

Hehe.. one guy had one of those.

I continued to just take a 5-1/4 floppy, turn it over, and use the notch of the top disk to provide placement for where to punch the hole on the bottom disk. Cheaper that way.

(And I realize that @mongo, you were talking about 3.5" disks and the drill press... very cool! )

I was jazzed when I got a Commodore 1581 drive that used 800K 3.5" floppies... it was great with GEOS.

 

[CuriousG]

Oh the memories! I used a drill press to convert 3.5" disks to high density.

I used a soldering iron. Wasn't pretty but it got the job done and ruin the tip.

 

[arcus]

What was that program that would try every combination of loading DOS drivers into lower or extended memory to get the most optimal fit (and still work)?

Man we had fun with that program.

QEMM!

 

[arcus]

I believe I may have used WinGate as well...

Same! I managed to find some used 3com ISA 10/100 cards and ran a cable outside between two rooms, so I could share my dial-up connection with my sister. USR was rock solid on flaky phone lines.