Space Shuttle Grouting

[Doctor Drivel]

[Andrew Gabriel]

In my 12 years working there, I only recall one incident of a machine bursting into flames spectacularly enough that it stood any chance of making a film (but sadly it wasn't being filmed at the time). More usually what happened is that a burning smell would appear in the computer room, completely diffused throughout by the aircon. People would walk around looking for the source but normally never find it, and you had to wait for someone to start moaning that a system had stopped working before you actually identifed the cause. In most cases it was actually some peripheral, normally a VDU, or occasionally a disk drive (washing-machine sized, not your little 3.5" drives of today).

[Doctor Drivel]

..and held 250 Meg on a Winchester drive and everyone went, wow!

[s--p--o--n--i--x]

Oooh! I had some brand new 8" drives, still in their boxes! Haven't a clue what I did with 'em though..

sponix

[dennis]

They held 5M on the fixed disk and 5M on the removable cartridge.

Before that we used drums that held about 5M and needed a screwdriver to unstick the heads if you left them stopped for too long.

250M on a disk is far too modern.

[Andrew Gabriel]

I have a disk pack from a CDC drive, with 10 14" platters (plus a guard platter top and bottom). Depending on sector size and the like, they are something like ~300Mb unformatted. These are about 2 generations before Winchester drives.

I spent a year or so designing a filesystem garbage collection utility for various disks back around 1985, and I liked to test it on these because I could take the cover off the drive mech and watch the heads for unexpected seeks (which slow the procedure down, so you try to avoid them). My pack is actually a "CE" pack (used for aligning the heads in the drive) which a field service engineer had head-crashed (a damn expensive mistake to make on a CE pack back then).

I also have a single 14" platter from these.

We still had these around for support purposes, but they were long obsolete in new machines when I started in 1983. A source of many nasty industrial injuries when people tried moving them while the drum was still spinning and overlooked the gyroscopic effect and the enormous length of time they took to spin down. Some customers were using them into the late 1980's because their realtime applications couldn't be transfered to run on disks with seek times.

[Doctor Drivel]

That's what I said. Whe they saw 250MB they went Wow!

[Bob Eager]

The first UNIX system I used had two 2.4MB hard disks. Both exchangeable!

[Andy Hall]

.... and you *definitely* wanted to avoid swapping....

[Doctor Drivel]

I recall using a DEC 10, the size wardrobes, with 150KB hard drive. It managed 200 users, and quite well too. It needed a permanent systems admin man on it to keep it up to scratch.

I have seen small tower servers using Novell running 1000 users plus, all using windows and no one permanently managing it.

[Bob Eager]

Couldn't help it, really. Memory was even more expensive...

But the swap area was a fixed size, created by making the file system (just one) smaller than the physical size of the disk.

The idea of 'partitions' for different file systems came later, and was in fact a botch answer to a design flaw.

[Andy Hall]

IIRC, on some kernels you had to compile in the swap area size as well and the two had to line up to avoid a nasty panic and a corruption

I didn't know that. I guess I started using/administering Unix systems with some of the first microprocessor based ones in the early

80s. These were mainly v7, System3 and 5 later. All had separate partitions as I remember. What was the design flaw? I guess it predated this period.....

[Bob Eager]

Yes...not sure about v6 (which was where I started in 1976). I'll take a look at the source sometime...I have it here!

Yes...basically, all disk addresses (sector based) were 16 bits, inside and outside the kernel. So the maximum disk size was 32MB (0.5K sectors). OK until the RP02 disks we had (20MB) were upgraded to RP03 (40MB). This meant that the raw device couldn't 'reach' the whole disk.

The solution was to split the disk into partitions ('virtual disks' in a sense), each of which didn't exceed the limit. Then the only thing that needed to do >16bit arithmetic was the disk device driver.

There was a lot of post-justification, saying that it was a good idea to split the disk up anyway (and that's true). But the original impetus was the 16 bit limitation.

The one system call that had to change was 'seek' since it used an int and not a long. In order not to break anything, it was supplemented (rather than replaced) by a 32 bit version, and that's why it's called 'lseek' to this day...

[Andy Hall]

I can vaguely remember an issue with device drivers having had some systems with 30 and some with 80Mb drives. Now it all makes sense.

Ah... I never understood why that was....

[Doctor Drivel]

screwdriver to

That says it. A Systems Admin man. Some fool called Capitol thinks you know all about heating and water systems. Maybe he got the word "system" mixed up. And Plowman in the mental system confused him too.

[Doctor Drivel]

A book by Dietel, now out of print, gives all that history. UNIX was a pun on Multix, a Honeywell derived OS.

[Doctor Drivel]

Attempted to.

[Doctor Drivel]

Another lunatic.

[Andy Hall]

I know that you're a plonker, but don't be a total one. I've written device drivers, network code and a broad assortment of other Unix stuff. I've never worked in a plumber's merchants though.

[Doctor Drivel]

Oh my God. The Systems Admin man spaketh.

You would have liked to, systems admin people always dream of that.

You would like to wouldn't you!