Misunderstanding Computers

Why do we insist on seeing the computer as a magic box for controlling other people?
人はどうしてコンピュータを、人を制する魔法の箱として考えたいのですか?
Why do we want so much to control others when we won't control ourselves?
どうしてそれほど、自分を制しないのに、人をコントロールしたいのですか?

Computer memory is just fancy paper, CPUs are just fancy pens with fancy erasers, and the network is just a fancy backyard fence.
コンピュータの記憶というものはただ改良した紙ですし、CPU 何て特長ある筆に特殊の消しゴムがついたものにすぎないし、ネットワークそのものは裏庭の塀が少し拡大されたものぐらいです。

(original post/元の投稿 -- defining computers site/コンピュータを定義しようのサイト)

Tuesday, December 18, 2018

High Cost of Replacement for Once Inexpensive Solutions

Someone in a Facebook group I lurk in devoted to "vintage {Computers | Microprocessors | Microcontrollers}" shared a link to a (three year old) story about how the heating systems of a bunch of schools in Grand Rapids had been operated under the control of an inexpensive Amiga computer for thirty years, which mentioned that the school system had scheduled a budget of more than a million dollars to replace it.

The Amiga system hardware would have cost no more than USD 5000 when they bought it.

But good engineers will tell us that the Amiga wasn't the only cost. 

The controls system was originally programmed by (get this) a high school student (who remained on-call to fix it for quite some time after). I suspect the hardware other than the Amiga was also laid out by students, teachers, and other school system employees, who probably did not charge the school system full engineering rates for their time. 

These kinds of controls systems aren't that hard, really, if you don't try to do too much.

The article notes that bugs have been found and fixed in the intervening 30 years, and new necessary functions added, as well.

This is what we call an organic solution.

If you ask me what they should be doing, they should be de-centralizing the system. And they should avoid using packaged desktop hardware and software like the plague. 

Why? Most of your packaged stuff are so big and complicated that the OS can't guarantee finite response times. 

Desktop is especially offensive because, essentially, the desktop operator has way too much patience. That patience means desktop systems tend to have so many bells and whistles that the entire hardware and software can be brought to its knees trying to pick colors appropriate to advertizing content out of multiple layers of style sheets in web pages.

(You thought you were impatient? Boilers in heating systems have much less patience than you do.)


Yes, if they go to IBM and Intel and Microsoft and ask for a replacement, it will cost a lot of money. Why?

First they have to analyze the existing system. That is really expensive. The rest is not so bad, but they will charge full engineering rates for their work -- and hopefully offer some sort of warranty. (Warranties are expensive!)

Wait. If that approach is taken, replacement could easily balloon into the multiple millions of dollars, because the market leaders will almost invariably try to do it using their pre-packaged systems.

Organic solutions that are shepherded along by competent engineers are usually the most cost-effective (and most effective) solutions because only the necessary functionality is added, only when it is necessary. (Mumbling something about so-called "agile" technologies.)

The problem is that "competence" is something of the luck of the draw. Certification does not really guarantee it'll happen. (Neither do large engineering fees replete with warranty and disclaimer.)

Why doesn't certification help? Think about those tests. Any test you can start and finish in a day is very one-dimensional. It can test a narrow set of knowledge and skills, but there is not enough time in a day, or even two to reach out and begin to test the branches that will be required in real-world application.

Spot-check, at best, which depends on the idea that honest seekers of certification will be studying the whole field, not just the narrow parts that will be tested.

And those taking the certification are motivated by that piece of paper. So they don't focus on the application, only on the part that is tested.

That's the excuse that testing agencies have for hiding the contents of tests. That would not be so bad, but the kinds of test questions that can be given in a certification test are limited to very non-real-world problems. So hiding doesn't really help. Teachers who know something about the field can guess close enough.

Students want to pass. Teachers want their students to pass. So they focus on teaching how to solve the non-real-world problems that can be given in tests.

This wouldn't be all bad, if those who accept certification then go out and apprentice themselves to someone who can guide them along and help them gain the real skills.

I think you can accept this much of my argument. The next step is one you'll likely rebel against.

Good computer and software engineers to apprentice oneself to are far and few between, and they are very busy.

Why?

Even now, we really don't understand much beyond the basics of computer science and software engineering. Sixty stinking years since the early big iron-and^glass, and we are still wasting time and other precious resources re-implementing old solutions.

Why?

We haven't got the basics down right, yet.

(See my posts elsewhere in this blog for bits and pieces of pointers away from the wrong directions we've taken.)

No. Let me get that out of the paranthesis.
We need to return to the technology we've left behind. 
Groups like the vintage FB group that sparked this rant aren't really just there to just play around. Young guys go join these groups for fun, they think, but these groups are where essential bits of the old tech are being passed on to a rising generation -- bits that would otherwise be lost.

Here's something else that was recently shared in that group, a video of a group restoring an IBM 1401 and some of the software that ran on it. (You may have to wait for a Grammarly or other (dispargement deleted) ad to see the video.) So they run the old compiler to compile and run an old example program to multiply and invert a large matrix of floating-point numbers. That sample program is not the real goal. The real goal is the technology they used to restore the hardware and software. Some of it uses standard solutions, some of it uses custom stuff built with modern parts, some uses new product of old designs that can still be found via the internet.

These things allow legacy systems that we don't know how to replace to continue to run. They help prevent hospitals and schools from suddenly going dark while the bean-counters aren't willing to wait for organic solutions to be put together.

It's very much useful.

But the IBM 1401 restoration is not nearly as useful as the other stuff these groups are doing, in my opinion.

(Again, look around this blog for bits and pieces of what I'm talking about. But I haven't had time to put the half of it up. One of the pieces I haven't been able to talk much about is real-time OSses. Some of the old real-time OSses still exist, just barely, with significant help from these groups.)

These groups are incubating ways to pursue the branches of tech that have been buried and left behind by the excessive competition of the marketplace and the industry of the last forty or so years.

Why aren't the industry leaders doing anything to fix this situation?

You've heard the old saying
Money is like manure. It's meant to be spread around, to help more little green things grow.
(Green being something of a pun on the color of US current tender.)

Many of the market leaders understand that.

(Pardon me for a momentary lapse of control, but I'm going to rant for a few paragraphs.)

A very few at the top, mostly those who have control of the accounting divisions, understand it differently.
Money is like pus. It tends to collect where the wounds in society are.
This is one reason we should not trust people who amass personal worth too much in excess of what it would require to retire at whatever age they are.

One million USD would be enough for a healthy forty-year old to retire reasonably comfortably right now, more than enough for a healthy sixty-year old.

I'm willing to stretch that a bit for active entrepeneurs, because they need to be able to make business decisions unhindered by the bean counters.

But I'm not going to place very much trust in anyone who has more than ten million in personal worth and is still trying to beat the competion. (... Implicitly, still trying to force his or her world-view on the market with the power such money can give.)

Well, I'm bucking the trends here.

But we need to go back to old CPUs, old systems, old toys that were really much more than toys, and find what we've lost.


2 comments:

  1. My employee and I still believe the suggestions are rules-based and a little one-size-fits-all. I confirmed this statement when I found this artificial intelligence app named INK for All.

    ReplyDelete
    Replies
    1. Not sure what you're referring to, and I'm not sure Ink for All has anything to do with what I'm talking about.

      Delete