I haven’t thought about that in years. My uncle was working there at the time; my aunt too, or maybe she was just visiting him. I vaguely remember them describing the experience. In any case, the “44th anniversary” part is what shocked me — I’ve been reminded that it happened nearly a half-century ago.

  Time flies like an arrow.
  Fruit flies like a banana.

I think I first saw that in a old Scientific American article about the difficulties in machine parsing of natural language. Lost in translation. A geek pun to me, since translation also refers to coordinate transformations: translation, rotation, scaling, skewing, … preserving topological equivalents across gross distortions, no tearing allowed. Ants. Moebius Strip II.

I sometimes envy those who have a sense of continuous, unbroken time. I imagine that for them events are like beads on a string, pages in a book, frames in a strip of film. Mention to them some event and they can provide context, cause and effect, sliding their finger along the curve and saying “see, this happened before, and this happened after”. The people who can always quickly answer those “where were you when such-and-such happened” questions that I hate. Steel Beach.

Like the “one, two, three, many” number concept attributed to some primitive tribes, time to me is only “today, yesterday, a while back, and a long time ago”. Time lies like pieces of a shattered mirror beneath my feet, glints and reflections, partial images fractured along random edges. Memory is a kaleidoscope filled with broken events, ever changing in random juxtaposition, tumbling in disconnected chaos. Mona Lisa Overdrive.

It is totally worth it. I’ve on several occasions, inadvertently as well as intentionally, pulled a stunt like this.

For example: once long long ago, when I was a teenager or just after, some friends and I were cruising around looking for some mischief to get into.

Me: Let’s listen to some Pink Floyd … clicks on the radio …
Radio: … Dark Side of the Moon …
Others: … get quiet and stare at me …
Me: WTF? What’s wrong?
Others: How’d you know?
Me: Know what?
Others: Pink Floyd. On the radio. You just turned it on.
Me: … smirk …

Internally, this freaked me out (keeping with period slang). You see, I actually had a Pink Floyd cassette in the player, and that’s what I’d meant to play.

It was just coincidence that Pink Floyd was on the radio. That I honestly didn’t catch on at first that it was the radio playing made me appear so nonchalant about it all, which is what really got to them.

image: Psychic, xkcd

Background

This was back in the day when the numerous DEC (later bought by Compaq, then by HP) minicomputers we had ranged in size from dorm refrigerator to 6-foot high, 10-foot long cabinet (not counting the storage systems). The unit of repair was usually a “module” — a circuit board that plugged into a backplane, or a disk HDA, that sort of thing.

We had service contracts with 4-hour (business day) response time. At the time, DEC had a real Field Service office in Jackson and a logistics hub in Memphis. Typically, when a service call was required, a FSE from the local office would show up with one or more “kits” containing related modules, and swap out the module most likely to be failing. Generally the first one or two modules tried would solve the problem.

If the local office didn’t have the right module (or the module from the kit was DOA), they would get it (or a whole kit) sent overnight from Memphis. In one or two extreme cases, they had drivers leave from Jackson and Memphis, meet halfway, and return.

The local guys were great. Also, once they realized I knew what I was doing, they’d trust me enough to just drop off parts, let me fix stuff overnight, and pick up the bad parts the next day. Good for me (no daytime downtime for a 1000 users) and good for them (more time for customers that needed it). More on this later.

However, the logistics guys in Memphis had had a really bad year. Way too many DOA parts in kits, or out of stock parts which they had to get shipped in cross-country.

The Survey

The infamous Customer Survey covered all aspects of support: local office services, logistics, software support, etc. It was a fairly long mark-sense (pencil in the circles) form, broken into those various categories. I gave the local Field Service office high marks (probably all 9′s and 10′s), but zinged Logistics as they so well deserved.

The Concerned Corporate Suit

Why DEC decided it was important enough to send a Concerned Corporate Suit (CCS) to visit me personally, I have no idea. Perhaps they had cloned a bunch of ‘em and needed to give them something to do.

In any case, DEC was well known for two things: first, generally excellent engineering; second, being so bad at marketing that they couldn’t give away ice water to someone dying of thirst in the desert. I understood why much better after this meeting, assuming marketing and probably most management (or at least the policy makers) were CCS’s.

Anyway, the CCS had come to explain to me the ramifications of my survey responses.

Responsibility Without Authority

Back to the Survey … now you’d think that this would make the local guys look good and that someone in Memphis would get his ass chewed a bit. But you see, that’s not how the survey worked. The FSE had a good reason to look nervous when he showed up at my office: the local office staff’s raises, promotions, and even continuing employment depended on the overall survey score. Despite the fact that the local guys provided excellent service, and had absolutely no control over the performance of the jokers at the logistics hub, their livelihood would be impacted by shoddy performance in Memphis — if you were honest on the survey.

Point #1 — responsibility without authority is stupid

Know What You’re Measuring … and Why

When you measure something, you expect variations in the results. Furthermore, you have to select the appropriate range of values so that you can see that variation. For example, if you’ve ever used a DMM, you have to select the correct range — trying to measure the voltage of your wall outlet using the 0-10 volt range is simply gonna peg the meter (or burn it out).

Satisfaction surveys are highly subjective and I would guess extremely variable, even if designed well and used appropriately. To have anything worth analyzing, you’d need to have some form of curve, maybe a bell curve, maybe not, but at least some shape that indicates that you’re picking up a sort of consensus with some outliers tapering the ends.

The Concerned Suit explained to me that Corporate expected all employees / centers to be rated 10, or (he implied) they were considered failures. Well, that’s easy to do if you set your criteria so low that only an earthworm or chicken would rate below 10. But if your 1-10 scale goes from “can drink a glass of water without drooling unless distracted” to “walks on water while chewing gum and juggling chainsaws”, then expecting a “10″ is a little outrageous.

But that was just my own naivete. The Survey wasn’t meant to be an instrument to evaluate and suggest improvements to customer service. It was actually a club to browbeat employees and a marketing tool (“We’ve been rated 10 by our customers for N years in a row!”).

Point #2 — if you expect to peg the meter, you’re not measuring anything

Gaming the System

I liked working overnight, and preferred to have downtime scheduled then because it inconvenienced fewer users (and limited the number of people interrupting me with “Is it fixed yet?” (see all these parts on the floor; do you think it’s fixed yet?) and other inane questions).

We briefly tried DEC’s Remote Diagnostic Console, which tied into DEC’s network and would proactively report potential and real problems to Field Service. Which was fine, except that it logged an “official” service call. Which meant that FS had to respond within four hours. Which meant that if we didn’t shut down and “repair” whatever was bad when they showed up (which we wouldn’t do arbitrarily, because we had our own service responsibilities to our users), they got zinged for taking too long to complete the service call.

So, the game was to notice any potential problem, informally call the FS guy and say I needed module XYZ123 for (let’s say) Wednesday night, and he’d drop off the part and “open” the call on Wednesday. If it was more difficult (that is, FS would have to do the diagnosis / work), I’d schedule downtime, and wait until then to log the call. Of course, for rare it’s-really-broken-and-we’re-down stuff, I’d log the call immediately.

In general, people will do a lot of whatever you’re measuring them with. Lines of code? Your programs are suddenly much longer. Number of calls answered? That single extended problem turns into multiple “minor” calls answered and closed right away. Ditto bug fixes, parts replaced, widgets produced, units sold, ….

Point #3 — the people you evaluate will game the system you use to evaluate them; choose your criteria wisely

My Solution

My discussion with the Concerned Corporate Suit covered points one and two pretty directly, and the idea of point three in general without ratting out FS. However, he was adamant that they expected 10′s.

Me: So you want me to lie on the survey, rather than giving you useful information?
CS: Yes.
Me: I see. I will no longer waste my time on your surveys, since you won’t address the problems I’ve indicated but will penalize the local office instead. If you want them filled out, send them to my boss.

They kept sending them to me, and I kept throwing them in the trash.

The easy fix would have been to set terminals to nobroadcast by default, and let anyone who wanted to receive calls to enable it at will, or permanently enable it in their login profile. But this was a case of an Administration’s technical solution to a social problem. So, losing the argument, I put an ACL entry on the phone execuitable to deny access to anyone in the student group.

Now, students are a determined lot, and they came up with a simple workaround. VMS, unlike UNIX-like systems, generally runs commands in the same top-level login process; and you can set various attributes for that process, including a name — set process/name="Hey, I'm in the AC lab!", for example. Heh. Then you can use the command show users (or specify a particular user) to see all the process names on the system….

So, at any random time, you could loop show users to see users displaying a Facebook-like status via process names or even carrying on dynamic Twitter-like conversations.

DCL allows abbreviations and command aliases, so avid users could define short commands to set and show processes; it wasn’t as tedious as seeing the fully spelled out commands makes it sound.

There’s really nothing new under the sun, just different ways of doing the same thing over again, albeit in fancier dress; a change in degree, not in kind.

Due to my own laziness, I had to renew my registration with Network Solutions [Wikipedia entry] instead of transferring to another registrar. Yes, I know I should have transferred it long ago.

Since that time 10 years ago, it has become possible to register domains with any of several registrars, and I’ve done so. It has also been easy to transfer domains among them. Except for NetSol. When I’ve transferred non-NetSol domains, it has only taken a few minutes or hours from initiation to completion, mostly dealing with approval emails and the time for the automation to complete.

However, when I transferred my one other NetSol domain, it took nearly a week to complete, including refused transfers, phone calls, and smarmy email trying to prevent me from transferring. Based on that experience I should have begun the transfer early, but I’d waited until too close to the expiration of lohnet.org; I didn’t want to risk the domain falling through the cracks in the middle of a transfer.

You have to initiate a transfer with your current registrar. It was almost impossible to find out how to transfer a domain away from NetSol, but of course there were plenty of informative links on their web site on how to transfer to them.

After finally discovering how, and jumping through the proper hoops to initiate the transfer, my new registrar kept getting rejection notices from NetSol whenever it tried to complete the transfer. That took actual phone calls to NetSol to resolve.

Then I started getting smarmy email messages from NetSol wanting to know why I was transferring, was I really sure, etc. The message that informed me the transfer had actually been accepted also informed me that NetSol would sit on the transfer for five days just in case I changed my mind, The only link in the email appeared to be a link to automatically cancel the transfer. It took another phone call to find out that the same link could be used to say go ahead and transfer now.

Next year I’ll pay more attention and begin the transfer early.

This turned into a brief rant about NetSol rather than history. Oh well.

Jeff Moser, in Computing History Matters, said ‘I had always assumed that the “core” referred to in the error message was an adjective for “main,” but when I saw real “core memory,” I realized that I had misunderstood the term for years.’ It’s an interesting overview; you should check it out.

However, that particular comment reminded me of yet more stuff piled in the closet (sigh)…

The image links to a photo set showing the details of a memory board from an PDP-8e, a 12-bit minicomputer, I once worked on. I rescued this card and the front panel when the system was scrapped. The card contains 4K words (12 bits each) of memory. I’m not sure if any extra bits were included for error correction, and I’m not gonna count all the little donuts to find out.

Flickr set: Zenith ZFL-181-93 Laptop