Section 74, WWS; Selectric innards (oxen!) <-- Re: i18n/Mac... (Lon

{Message summary: Detailed replies to Peter T. Daniels' recent message,
partly in response to my comments about Section 74 of WWS, "Analog and
Digital Typography". Maybe from about 20% to 60% through the whole message
is a diversion that can be skipped, and probably should not have been
posted. I tried to make it easy to follow, somewhat entertaining, and also
easy to find where it ends. --nb}

On Sat, 26 Mar 2005 21:39:16 -0500, Peter T. Daniels
<grammatim@...> wrote:

> Actually, as the footnote points out, it's a talk presented in Princeton
> in 1988. Ephraim Isaac asked me specifically to talk about computers and
> writing.

Woops! My mistake. I should slow down (more) and be more scholarly. (Just
noted a simple typo: "Hammod" for "Hammond"; anyone familiar with
typewriter history knows about the Hammond.) For the record, scalable
fonts, at least, if not outline-defined fonts, were known in 1985.

> (I got him to invite me to the other session of his conference to talk
> about the origins of writing, too.)

Imho, something you're very well qualified to talk about.

>> Naturally, I was extremely interested in reading that Section. To put
>> it not too gently, the author didn't understand the esssential basics
>> of his topic. It might have been written by an art critic.

> Thank you! That's basically what I was aiming for.

! --nb

Some art critics, those I consider the best, make definite cultural
contributions, help to educate, and can be understood by many. There are
others.

>> Offset (lithographic) printing is significantly different from
>> mimeograph.
>> I know enough about both technologies to be confident of stating that.
>
> I'm, once again, not interested in the technology.

Your mistake. That is no excuse for making factual errors.

> The common term is "lead type." It's not "lead-alloy type."

I've set a very modest amount of type by hand, in a composing stick; just
a few lines. I know of the alloy as "type metal".

Justification is a nuisance, and the spacebands in Linotype were a
brilliant solution: Simple pairs of wedges. I never learned why the molten
type metal doesn't leak.

[In rereading Section 74, I just spotted another "whopper", a spectacular
error: Saying that a [mechanical] adding machine is an analog computer! I
can imagine hoots of derision among computer historians, because that is,
by all definitions, so *utterly* and exactly wrong. (But, please see the
next paragraph.) I am the resident expert on analog computers, as far as I
know; I repaired* an awesomely-complicated mechanical analog [gun]
fire-control computer as a Navy technician, and have written articles on
the topic for another YahooGroups mailing list (Howthingswork). I had
formal training in analog computers in the Navy, and can speak with some
awareness about electronic analog computers. I know what I'm talking
about. I have had decades to ponder the philosophical differences.
*Only one loose screw, but getting at it should have taken a day;
super-tight critical screw took almost 3 days to loosen, using a
four-foot-leng screwdriver with extensions.
<http://dcoward.best.vwh.net/analog/ford.htm> gives the idea. One I worked
on was only slightly different.
Power consumption was normally nowhere near that much.]

However, your next phrase (about the distance a gear wheel moves) is
interesting, and, (surprise!) I agree with it! All present-day digital
devices are composed of "analog" components; what matters, however, is
whether their routine and normal use involves continuous, stepless change
(analog), or only some number of discrete states (positions, numbers,
etc.), in which the analog transition between states is not part of the
way the information is represented. By convention, the latter (discrete
states) is now all but universally referred to as "digital", because it is
so phenomenally well-suited to representing numbers. Quantum computing is
a different matter; no analog, continuous transitions, afaik.

A related term is "quantizing": Fitting a continuously-changing quantity
(time, temperature, pressure, voltage, etc.) into a representation that
doesn't have a continuous form; the latter exists only as one of many
"steps" (Huh?)

It's simpler than you think: When you read a thermometer, you quantize an
analog quantity (continuously, smoothly-changeable column height or
pointer position) into a number, generally one or two digits (and three,
in the USA) for ordinary air temperature. The temperature is a number, and
the scheme is set up so that there are enough numbers to represent,
closely enough, for day-to-day use, vhat the analog quantity is.

Nevertheless, you routinely quantize a temperature reading of a room (or
outdoor) thermometer when you mentally think of only a few digits. That is
good enough for anyone.

If that thought didn't scare you off, reading a thermometer and expressing
its reading as whole numbers is doing an analog to digital conversion, in
this case, mentally. (If you're reading this, and badly confused, that's
my fault.)

Remember: Analog quantities change smoothly, without steps. Digital
quantities are constrained to have only specific forms, usually considered
to be represented by numbers. They change by "jumps". They are said to be
quantized. When you read a thermometer as described, you quantize the
reading, as I said.

A slide rule is an analog computer. You can place two rulers near each
other, parallel, to add inches and fractions (or millimeters).

Mechanical sophistication is not a prerequisite to the definition of an
analog computer, nor is cost of materials.
The [Vannevar] Bush Differential Analyzer was an analog computer, one of
the most famous.
Comdyna still makes electronic analog computers. (!)

More generally, however, what really matters is not the numerical aspect,
but it is that information is represented by a collection of individual
items (often, bits, in electronics), or positions (mechanical decimal*
calculators and adding machines). Those items work right only when they
are all in one or more of their steady states, and *never* in between.
*Historically, also Sterling currency, too. Was fun.

To represent the information in question with the required faithfulness,
the makers provide as many items (gear wheels, or electronic bits, for
instance) as needed. That degree of faithfulness is numerically described
as precision. Do you need to handle larger numbers for currency (Yen, Lira
before € (Euros), others)? You need more digits in your adding machine.
Need better color? Go from 8 bits (urgh!) to 24. Is 16-bit CD audio not
good enough? Try 24 bits. (Some do.)

However...16 bits not enough to represent all the needed glyphs? Hmmmm. No
talk that I've heard of, about 32-bit Unicode. Extensibility was Very
Wise. (A little distraction, for you.) However, the "U+nnnn" (4 hex
digits) is now "U+nnnnn" (5) when needed.

(Accuracy and resolution are related concepts, but, although important,
this is not the place to elaborate.)

This fundamental philosophical difference between discrete and continuous
is important, and useful, to understand. Concerning typography, if you
have enough image-area bits (rows and columns), you can create glyphs that
are "good enough". IIrc, the Linotronic runs 2540 bits/inch, suspiciously
like 100 bits/mm, if my quick mental arith. is on-track. That's a nice,
fine grid.

(Btw, the outlines that define glyphs in modern outline fonts are surely
more "analog" in character, even if the control points that define them
are constrained to a (finely-spaced) digital grid (afaik!) I truly doubt
that in a font file, they exist as "analog" curves; they exist only as the
data needed to create those curves. However, the way fonts are dealt with
by a computer imitates "real-world" analog-like processes. Interesting
thought that just occurred to me (and was amplified upon in the rereading
befole posting).)

[nb]

>> The Selectric typewriter is not electronically controlled. Technically,
>> it is quite remarkable. It has an electric motor to operate it. If the
>> speed were right, it could be powered by a foot treadle, like sewing
>> machines.

> How does it know which of the 80-odd facets to apply to the paper?

I'm happy you asked; curiosity is often good. You have presented a real
challenge -- to describe the innards of the Selectric in non-technical
terms, so that everybody can understand, and concisely enough to not
(more) flagrantly violate the prime intent of Qalam.

P.S.: I might have omitted some detail, but have made this long,
attempting to make it more understandable. For variety, I brought in some
horses, only to Google for a good illustration and change to oxen. I'm
revising the text to create an easily-visible indication (bunch of shorter
lines) at the end of this explanation, but I tried very hard to make it
thought-provoking and somewhat entertaining, being careful of technical
lingo.

----------------------------------------->8======================
(Left-handed scissors... (Arabic?))

OK: The "golf ball", a very-popular nickname for what (iirc) IBM might
call the "typing element", plainly rotates; it also tilts, to select the
proper band of characters. What we are concerned with is how the machinery
positions the ball, tilt and rotate. IBM's terms, too.

Before going on, I should say that the Selectric is a masterpiece of
mechanical innovation, imho. Its innards are dramatically similar in basic
principles to commonplace electronic devices, in some significant
respects. (Technically, many of its mechanisms are binary, and it contains
two mechanical digital-to analog converters. Don't let that bother you.)

Each key (almost every one, if not all) has its own, associated
lightweight metal bar that has a spring, and that spring is normally
"tight", "tensed". These little bars are almost horizontal, maybe exactly
so, and run (part way) between front and back. They sit side by side.

When you push a key, you operate a little latch. Like a door latch, you
free up the little metal bar so that the spring can make it jump to a new
position. The fun begins.

The electric motor (or foot treadle! I wonder whether anybody...) spins a
shaft; we techies would call it the main driveshaft. There are several
clutches on that shaft (think "stick-shift" car clutch) that are sitting
disengaged, waiting...

Those clutches, when they engage, connect the driveshaft, so it moves some
specific part of the mechanism. (New typing line? Trip that particular
clutch. The ball goes back.) Car clutch: You take your foot off the pedal,
and the engine turns the wheels.

OK, once the little bar (we would call it a code bar; don't let that
bother you) is released -- unlatched, it trips one or more clutches (like
taking your foot off the pedal in a car). The mechanism that is now
moving starts the process of positioning the ball.

However, this not being a book*, I'm getting ahead of myself.
*P.S.: Pamphlet?

Every key's little bar, code bar, is different. It has a[n] unique pattern
of bumps (or no bumps) on it. For each particular bump (say, the second
one, for tilt-position code), there is another bar, which runs all the way
across the machine -- all across the keyboard.

So, you have two sets of bars: The little ones (perhaps 3 inches / 75 mm
long, very roughly) for each key, and the other, long ones, which run
crosswise. As to the latter, there are something like ten of them; I don't
remember. (10 bits...)

Similarly to what happens when you poke a key on a computer keyboard (*in
general*, *not* in detail!), when you poke a Selectric key, the little
bar's unique pattern of bumps and missing bumps defines a combination of
bits -- yes, binary, 1 or 0. Surely you remember the bars that run across
the machine. Well, guess what. When the little bar is unlatched, only some
"cross" bars are moved by the bumps. Missing bumps, naturally, don't move
their corresponding "cross" bars.

OK: Now we have only some "cross" bars operated (think binary 1s) and
others are not (0's). I don't remember details, but the "cross" bars
can't directly operate the ball positioner mechanism. Hang on, I'll be
changing my spots again, calling them "code bit" bars...

What the "code bit" bars do, instead, when a bump operates them, is to
move a piece of metal* "into the way", into a space. That space is between
a powerful moving piece (remember the clutches?) and individual other
movable pieces, one for each "cross" (code bit) bar. Those other pieces
are what operates the ball positioners. If a bump sets a bit by moving the
"cross" (code bit) bar, then that bit in the ball positioner gets moved by
a standard amount. If a bump is missing, its corresponding metal piece
(interposer) doesn't "get in the way", and the corresponding part of the
ball positioner doesn't move.

*Technically, an interposer, something never mentioned in the old list of
(seven?) basic mechanisms

Good Lord, I'm testing Seshat's and others' patience...

Okay. I hope I've established that poking a key sets up a particular
(binary, digital!) combination of parts of the ball positioner. According
to which specific combination exists, the ball will move to a
corresponding position. Remember, one positioning mechanism for tilt,
another for rotate. Btw, so far, nothing technically unusual.
Teleprinters (Teletype [tm]) had keyboards like this.

[Horses and wagons]
(Surprised?)

So, when you have a big wagon, a team of horses to pull it, and wide roads
(I assume!), how do you arrange it so all horses pull one wagon? Maybe you
have the horses in pairs, side by side, and a very long pole, or other
common hitch between; that's not what I have in mind.

Ever hear of a whiffletree? {Googling for images}...
{Back} Aha.. Just one good one, before patience ran out, looking at many
dozens of nice room interiors...
<http://cougar.slvhs.slv.k12.ca.us/~pboomer/physicslectures/machines1/whiffletree.jpg>
Kindly try a look? Top-down view.

OK, so it's oxen. Fewer letters to type.

Please note the arrangement that attaches the oxen to the wagon (that's a
very wide wagon, btw!).

OK, you now see the basic scheme used to position a Selectric golf ball.
If one ox moves forward one standard ox-step, and there's no slack, the
wagon move forward by a standard amount, actually one-quarter ox-step, but
that reduction isn't important. Two oxen move the cart twice as far when
they take one standard ox step. Of course, the cross-beams have swivels --
pivots -- at their ends and middles.

[Side note I can't resist including: That whiffletree linkage is used in a
category of mechanical analog computers called linkage computers. It adds
(and subtracts) mechanical analog movements that represent quantities.
Lots of whiffletrees:
<http://dcoward.best.vwh.net/analog/libra.htm>, right column of images,
upper]

Selectrics and equal-opportunity oxen differ somewhat. You'll easily see
that the cross-pieces that pair up the pulls have a common point at their
middle. In the Selectric, that common point is usually not at the middle;
some oxen are more equal than others (Orwell...)

Nevertheless, The Selectric ball positioners are very much like
whiffletree linkages. The bumps --> code bars --> interposers (and
finally some power in the picture) are like the oxen -- they pull the
wagon, to tilt and rotate the ball.

A Selectric extends the whiffletree to more oxen/bits. As I remember,
there might be as many as ten bits, but surely not 1,024 ball characters!
there are many "illegal" combinations of bits, philosophically much like
"no defined character" in Unicode.

[Here, it might be worth noting, that internally, Selectrics have their
unique code; it's nothing like ASCII, although it is weighted binary, but
not 8-4-2-1, either.]

Naturally, all the mechanism described is inside the main part of the
machine; it doesn't go across the page with the ball. Something else I
think I should skip any clear and painstaking explanation of, is how the
mechanism's motion reaches the ball. Concisely, those thin, taut
copper-colored metal bands run over round pulleys, most of them mounted on
the frame.. One end of each band is fastened to the ball carrier. One of
those round pulleys, however, is an exception -- it's "at the wagon" --
The pulley there moves; it tightens and loosens the band. On the ball
carrier are two springs (one for tilt, one for rotate) that pull on the
other ends, and keep the bands taut. As the "wagon" moves, those springs
stretch and relax, and that motion is what positions the ball. Clever. I
lied. I described it, anyway.

A lesser matter is that these mechanisms only get the ball close to its
desired position. Those teeth on the bottom edge of the ball act like
wedges, in a sense, to exactly set the rotate position. I've forgotten how
tilt is exactly set, but it's the same principle.

I'd love to know whether this explanation could be followed; surely hope
so! Little doubt it contained some surprises. Added challenge to do it
with no, or minimal illustrations. I do this fairly often for the
Howthingswork list (will abstract for them), but more
concisely/technically. No mention of cams, here, for instance.

OK?

THE END... OF TALES OF OXEN, WHIFFLETREES, AND
SPINNING GOLF BALLS THAT GO "PLOCK!"

Good grief, only about 60% through?
Should have broken this up into sections, and
posted separately -- er, maybe kept it much shorter...

Skipping some lines...

and more...

and more...

Resuming my replies:

Pipe organs bult some time ago have mechanical binary memory for
registering combinations. Casavant used bellows, organ air, to operate its
capture combination action. Each bit was a small pivoted piece with a
spring to ensure it didn't get stuck part way. It took a considerable
amount of development, I'm told, to create really good electronic
equivalents.

> What about the next generation, the typesetting machine, with variable
> spacing, automatic justification, columns, etc.?

Proportional spacing was originally mechanical. (A correcting Selectric
(must have been prop.-spaced) even had a mechanical memory for the last
few dozen characters typed, permitting repeated proportional backspacing
for lift-off correction, iirc.)

Automatic justification dates back to the Paige typesetter, a phenomenal
machine, development supported by 'Mark Twain'. The Linotype's paired
wedges (spacebands) were a delightfully-simple approach; Paige apparently
had a much more sophisticated scheme.
<http://www.twainquotes.com/paige.html> Not necessarily the best ref.;
Google ranking.

The Photon phototypesetter (at least the earlier model[s]) used
electromechanical relays to do what would now be done by a computer
program, essentially, and as far as I know. (See an engineering book by
Higonnet and Grea; that book was created while doing very sophisticated
design on the Photon.) (Relays are electrically-operated switches. Apply
power to the coil, and the magnetism that creates attracts a "flap" to
move the switch contacts. Rotary-dial-era phone exchanges used them by the
tens of thousands.)

Eventually, electronics took over all the functions that it could.
Stored-program digital computers (the longer, accepted formal name) are
just too versatile to not be used.

>> Phototypesetters used (precisely-timed) flash lamps. (I want to see why
>> I noted that...)

> The one I used (Varityper 6400) used a CRT, not flash lamps.

Some early phototypesetters had a disc with photographic transparent
glyphs very carefully located near its edge. The disc was changeable, to
change type faces. (Families?) When the desired character was exactly in
position, the flash lamp fired, exposing the image onto photosensitive
stock. These were outdated quite some time ago, afaik; very unusual sets
of lenses (think telescope, but quite different) offered different point
sizes. These were available pathetically cheap, surplus, not too long ago.

I'm woefully ignorant (really!) about modern phototypesetters, but it's
very believable that they used exceptionally-high-quality (monochrome)
CRTs to create the glyph images (they might still use them). Monochrome
CRTs are capable of astonishingly-good resolution (think fine detail);
they, and their associated equipment, are not cheap. (Color CRTs have to
satisfy two philosophical opposites: They have to keep primary colors
completely separate, yet spread them evenly over the full viewing surface,
with no easily-visible boundaries. Interesting? Magic? ("Any technology
sufficiently advanced is indistinguishable from magic"). It's not magic;
it's shadows! See why I also like technology?)

> It was definitely the last bit that went into the book.

Good!

> I noted that people _don't_ save drafts. They just do their revisions in
> the files.

I once learned, the hard way, that (even 8-inch) floppy disks are
incomparably cheaper than the possible value of a complete archive. A
flawed computer program was giving real hints of solving a very difficult
problem. Fixing the mistake in the program (and writing over the old one!)
put me back into limbo again. No hint.

I plan to re-read the WWS Section again, but as usual, my enthusiasm has
put me off-schedule, and the sun has been out. Looks lovely.

> I don't need to know how they work, just that if I want an accented
> letter, I type that key first.

Surely! I agree. "Dead" keys: Actually, it's no more than a missing bump.
Just that simple.

>> If you were using a keyboard layout for another language, you'd want to
>> know about the corresponding variety of keys?

> The keys have labels on them.

Complete behavior of Caps Lock is not obvious (nor is Scroll Lock!),
especially to a typewriter user. (What's "Alt"? "Ctrl"?)

But you can't tell how to use them just by looking; you need to learn.
(I've remapped my keyboard to International, although gone back to
ordinary. Had to re-learn to use quotes, for instance. Dvorak layout means
one has to prepare at least two, and better, four new code charts: Shift
pair, and Alt-Gr pair. I used SC Unipad...)

> Around 1977, I went with my mother to Tytell Typewriter to see their
> stock of foreign-language typewriters. While chatting with Peter Tytell,
> I showed him the keyboard of the Assyrian Dictionary's custom
> typewriters and the changes I'd make to turn it into a Semitic/general
> linguistics typewriter -- and my mother told them to make me a custom
> Olympia. I still have it, of course, but I doubt the ribbons are
> manufactured any more.

Very nice tale. Wish I had more interest in languages, back then.

Ribbons: Take heart! Google on [olympia typewriter ribbons]. Caveat
emptor, as ever.
Son of a gun, I find that even the ribbons for my dual-encoding ('8859-1,
and (probably) "DOS" (Codepage 437)) Seikosha MP-1300AI are apparently
both available and even affordable.

>> Surely, for instance, you do know how to use a floppy disk, and it's
>> very likely you know how to set a floppy to prevent (or enable) adding
>> or replacing data on it. Mac made using floppies even easier, from the
>> beginning.

> What does that have to do with knowing how a computer works? That's like
> (as you analogize below) knowing that I need to depress the clutch pedal
> before turning the key on my Saturn.

There are, plainly, two different kinds of knowledge that seem to be
overlapping and becoming confused. Knowledge of internal principles of
operation, for the most part, is not needed by those who operate most
devices. (Refineries and modern small private steamboats are some
exceptions, but we're referring to devices usable by the general public.)
I feel 5% brain fade, but what we are hammering on you about is knowledge
of how to use, not internal operating principles. It's utterly irrelevant
to know that the read/write heads in a floppy drive are pressed into
contact with the surface, and that a rigid ("hard") disk's heads do not
touch, but operate inconceivably close to a very smooth surface. Other
than knowing that "hard" disks are somewhat delicate, those bits of
information are irrelevant.

> The Command key has no use of its own at all. It's there to add
> functionality to the other keys. And you didn't mention the Option key.
> (And some programs even use the Control key.)

However, I dare say you know many, maybe even all of the ways to use
those; I call them "qualifiers". Apologies for my ignorance of Macs;
again, they are humane.

> I have (having created some of them) all the fonts used in WWS. I don't
> know that having Unicode makes the exotics any easier to use -- I just
> select the font I need from the Fonts menu. I then either select Lloyd
> Anderson's specific keyboard, or I go to PopChar to select characters by
> shape. (His Syllabaries keyboard is a work of genius.)

Imho, an interesting insight into how you work. However, considering that
you send e-mail to a publicly-accessible mailing list, as recent
experience has shown, you can no longer completely work with total success
in that environment. You need to learn what characters are compatible with
(at least) Latin-1, a.k.a. ISO 8859-1; with the advent of the Euro,
'8859-15 is a better idea. I need to check on '8859-16, btw.
This is what we have been hammering about!

> All you do is look at the typewriter's keyboard, and you know all there
> is to know about using it.

Yes (mostly), and no. Shift lock? Youngster thinks it's identical to a
modern Caps Lock key? Typewriter Tab key? Not quite. While the process of
pushing down a key to print a letter is easily understood by most of the
world's people, I dare say a young US native-born person might at first be
puzzled by the "dead" keys on "foreign" typewriters.

(But, then, we use diacritics as decorations. I love the tale, likely to
be true, of a rock concert in Germany. As the band came on stage, the
audience of fans chanted in unison the (partly-awkward, imho) German
pronunciation of Mötley Crüe.))

> See C. P. Snow, The Two Cultures. Written quite a long time ago.

Thank you, indeed. Partly to make up for the lifetime consequences of a
mostly-miserable single term at Princeton, I have emphasized the sci/tech.
side of my continuing self-education. (I wasn't ready for Princeton, and
they weren't ready for me, in summary. Didn't quite flunk.) Yes, I do
recall C.P. Snow as a worthy author. I'll try to borrow it; easier weather
helps.

> _Something_ is coming up in my front yard, but I hadn't seen my house a
> year ago, so I don't know what they're about to be.

Nice way to end a message!

Btw, I'm deferring re-reading §74 until some other time; need to get on
with my day.

Thank you, Peter; I'm tempted to say you give as well as you get. I got
over my fury days ago. Ref. to C. P. Snow is appreciated.

As well, my apologies to all who don't give a hoot about technological
philosophy.

--
Nicholas Bodley /*|*\ Waltham, Mass.
The curious hermit -- autodidact and polymath