Matrix Teeth to Magazine Channel to Character

image link-topic-sf0.jpg

1. Warning/Apology/Plea

The method of this Notebook hovers somewhere between "reverse engineering" and forensic reconstruction. What keeps it from being actual forensics is the fact that the information I need to complete it accurately does in fact exist - I just don't have access to much of it.

So, first, I would ask any reader who spots a mistake to let me know, and any reader who understands this subject to help me to understand it.

Second, I could very much use any or all of the following items or sources of information:

(1) Broken, damaged, or just plain worn out Distributor Bars of any type except 90-channel (that is, 72 channel narrow or wide, 34 channel side narrow or wide, or (rare?) 28 channel side; I already have 90-channel distributors on my own machines). Why broken parts? First, because I can't afford actual new-old-stock working parts, and second, because I don't want to take working parts out of circulation (there are few enough of them as it is). The purpose of having these parts would be to examine (and publish) the rail patterns on them - to verify that my reverse-engineering of these rail patterns is correct (or fix it if I'm wrong).

(2) Failing actual parts, clear photographs of these same distributor bars would suffice. It's harder to take a clear photograph of a metal surface than one might initially think, though.

(3) Any relevant Linotype literature prior to Rogers' Linotype Instruction Book (1925) or prior to the Linotype Catalog No. 22 (1925) or Linotype Catalog of 1905 (which I have). The same for Intertype prior to 1929 ( The Intertype) or the 1932 Intertpe parts catalog. In particular, early Linotype keyboard arrangement diagrams would be relevant (by the 1937 edition of Useful Matrix Information, the lowest-numbered diagram was No. 6; surely diagrams 1 through 5 existed). So would any material having to do with the introduction of the 72-channel magazine or the introduction of any of the side magazines. So would any discussion of the keyframe mechanism for any of these non-90-channel arrangements.

(4, long shot) It is unlikely that anyone has sitting around on a dusty back shelf any broken/damaged Linotype/Intertype keyboard or keyframe assemblies, or Linotype "Channel Chooser" (Model 28 on) or "Keyboard Rod Lever Slide Shifting Mechanisms" (Models 8, 14, 25, 26), but if by some chance you do, what's junk to you is research material to me. Even though a part/assembly doesn't actually work any more, so long as at least most of it is there you can understand a great deal about how it should work, and this knowledge is important to me.

(5) Even knowing the actual introduction dates of various features would help in tracking down information on them. I don't know the dates of introduction for:

These dates would probably differ for Linotype and Intertype. Note: According to Leonard Spencer ("Linotype Models" on, the 34-channel 11 1/4 inch wide side magazine was introduced on February 1, 1930.

If you can help in any way, please contact me .

It is likely that relevant information does exist in the US patent literature; I simply haven't tracked it down yet. (There were a lot of Linotype and Intertype patents.)

There's also a question of audience here. In the end, the "audience" for this Notebook is me; it's just a Notebook, after all, not a formal study or paper. But given the nature of the Internet, it is possible that others will end up reading it. The subject matter here is pretty detailed Linotype/Intertype technical information, so it makes sense to assume that a reader knows the basics of the Linotype fairly well. On the other hand, also given the nature of the Internet, some readers may stumble across this with very little Linotype background. Those readers might be better served first first reading some of the general introductions to the Linotype and Intertype in the Linecaster Bibliography Notebook. (If you got here somehow and don't even know what a Linotype is, then you're very lost; you might wish to start with A (Brief) Hot Metal Taxonomy.) In the end, I've compromised and included a little more information here on the path of the matrix through the machine than is necessary for an experienced operator/machinist.

2. The Tooth Chart

Here's a chart which lists all of the possible tooth combinations (as binary numbers; thus the whole table is sorted by tooth combination as a number) with the magazine channel associated with that tooth combination (for 90, 72, 28, and 34 channel magazines) and (for orientation, mostly) the character most usually associated with that tooth/channel position. It's probably best to print this out; it's a PDF format file. linotype-tooth-combination-comparisons.pdf

If you prefer, here's the original OpenOffice spreadsheet version: linotype-tooth-combination-comparisons.ods

3. Matrix Teeth (and the "90-Channel" Magazine)

[This section is repeated from the previous Notebook, The Matrix Tooth Pattern as a Binary Number . If you're already comfortable with counting in binary, skip that and start here. If counting in binary sounds a little strange, it might be best to start there first.]

If I pull out a copy of Linotype's Useful Matrix Information (in any of various editions), somewhere in it will be a set of tables entitled "Tooth Combination Chart." (In the October 1937 edition, it's on page 16. In the undated but probably 1950s edition which is black with a gold matrix partial profile on the cover, it's page 18. There are equivalent charts in Linotype Machine Principles (p. 274) and elsewhere.) The first of these tables lists the channels of the standard "90-channel" Linotype magazine (which actually has 91 channels, numbered 0 through 90). It lists the channel number, the physical width of the channel, the character which normally is to be found in the channel, and finally the "Teeth in Combination."

This listing of the "teeth in combination" for each channel position is at first confusing. It starts out:


Clearly, there is some pattern - but what is it?

The numbers, 1, 2, 3, 4, etc. refer to the tooth numbers. Linotype numbers the teeth on a matrix from 1 through 7 (there are 7 teeth on either side of each matrix, but since the left and right teeth of any pair are always the same this works out to 7, not 14, teeth per matrix; you just count one side).

(From Useful Matrix Information. (Brooklyn, NY: Mergenthaler Linotype Company, October 1937.) p. 14.)

So what they're saying is that the matrix teeth for the tooth positions identified are those which are present. Let's call that state "1". Matrix teeth which are not identified are absent; let's call that "0". So for a start, let's lay out all of the teeth and see which ones are present or absent. It starts out like this:

1 2 3 4 5 6 7 
0 1 0 0 0 0 0 magazine channel 0 = tooth combination "2" 
1 1 0 0 0 0 0 magazine channel 1 = tooth combination "1-2" 
0 0 1 0 0 0 0 magazine channel 2 = tooth combination "3" 
1 0 1 0 0 0 0 magazine channel 3 = tooth combination "1-3" 
0 1 1 0 0 0 0 magazine channel 3 = tooth combination "2-3" 
1 1 1 0 0 0 0 magazine channel 4 = tooth combination "1-2-3" 
0 0 0 1 0 0 0 magazine channel 5 = tooth combination "4" 

This starts to look very much like binary counting from left to right ("big-endian"), but there are some complications.

First, the tooth combination "0 0 0 0 0 0 0" (that is, no teeth at all) isn't indicated. Matrices with this tooth combination simply fall into the "quad box" (the Linotype term; Intertype calls it the "transfer channel quad box" or "quad tumbling box" {Sinclair. The Intertype. (1929) p. 238.}) This does, therefore, really count as a matrix combination (Sinclair indicates that it is used for logotypes, thin spaces, and other special characters). It doesn't show up in the "Tooth Combination Chart" tables because these matrices drop out before distribution, so they're never involved in getting put into a magazine channel.

Second, the tooth combination "1 0 0 0 0 0 0" simply isn't used.

Third, Ottmar Mergenthaler was a machinist, not a computer programmer. So he didn't indicate binary digit value; he just counted teeth. But this doesn't matter, since what is important is the pattern, and its recognition by the distributor bar - not the number it represents.

Putting it all together:

1 2 3 4 5 6 7 tooth number
1 2 4 8 16 32 64 binary digit placevalue (in decimal)
0 0 0 0 0 0 0 no teeth, falls in quad tray
1 0 0 0 0 0 0 tooth combination not used
0 1 0 0 0 0 0 magazine channel 0 (combination: "2")
1 1 0 0 0 0 0 magazine channel 1 (combination "1-2")
0 0 1 0 0 0 0 magazine channel 2 (combination "3")
1 0 1 0 0 0 0 magazine channel 3 (combination "1-3")
0 1 1 0 0 0 0 magazine channel 4 (combination "2-3")
1 1 1 0 0 0 0 magazine channel 5 (combination "1-2-3")
0 0 0 1 0 0 0 magazine channel 6 (combination "4")

So there it is. The tooth combinations are just counting in binary, from left to right.

That's of course a lovely pattern, but what does it do? The Distributor Bar answers that. Here's an illustration of the beginning of a standard 90-channel Distributor Bar.

(From Linotype Machine Principles. (Brooklyn, NY: Mergenthaler Linotype Company, 1940.) p. 270.)

Black lines indicate that a "rail" is present in the distributor bar to engage a matrix tooth, white space indicates that the rail is absent. The first part to the very left, which shows all seven rails, should be ignored - it just carries any matrix onto the Distributor Bar. The first detecting position is that marked "e 0" (magazine channel 0, which usually holds the extra 'e' matrices). If a matrix passing this position has only the "2" tooth present, it drops off. All other matrices which have reached this point have at least one other tooth present, so they continue on. In the next position ("e 1") matrices with only teeth 1 and 2 drop off; all others continue. This pattern continues through the end of the "90-channel" magazine (combination 3-4-5-7, which is "0011101" (left to right), which (again, reading left-to-right) is 92 (decimal). So the 91 channels of a "90-channel" magazine, numbered channel 0 through channel 90, correspond to tooth combinations from 2 through 92.

Note that this mechanism does also imply an order to the detection. Consider the fairly easy position for channel 2 ('t'), which has teeth with only tooth 3 present. In all positions previous to this which have in fact been encountered, there has been a rail for this tooth, so this matrix hasn't yet dropped off. But in fact this matrix will drop off in any rail pattern which is missing rail 3. So if, for example, the Distributor Bar order were somehow reversed to "e e a t", then this matrix would drop off at "a" (which would be wrong). The binary counting pattern assures that this cannot occur (which is why the exceptions to binary counting which will come up later are all the more interesting).

In case the whole concept of matrices dropping off the bar at certain rail/tooth combinations is still a bit murky, here are some diagrams to help. The first is Linotype's version, showing a matric for a lowercase 'l' as it encounters position 11 on the Distributor Bar (the diagram of the Bar, above, shows this location with an arrow):

(From Linotype Machine Principles. (Brooklyn, NY: Mergenthaler Linotype Company, 1940.) p. 270.)

Unfortunately, the drawing above, while accurate, isn't particularly clear. The equivalent Intertype drawing, reproduced below, is much clearer. It shows a matrix held on the distributor bar rails by its teeth, on the left, and the same matrix released at the appropriate location along the bar, on the right. (The matrix shown is a different one, 'v' rather than 'l', but the principle should be clear enough.)

(From The Intertype. (Brooklyn, NY: Intertype Corporation, 1943.) p. 272.)

(The illustrations, below, of the path of a matrix are mostly for those readers who might not be as familiar with the Linotype as they could be. This will all of course be "obvious" to an experienced Linotype/Intertype operator or machinist.) The matrices, having been delivered to the top of the machine, travel along the distributor bar (moved by long helical screws) and at their appropriate locations drop off, pass through a curved "channel entrance" section, and go into their magazine channels. The illustration below shows this, in a partially cut-away view from behind the machine. In it, the helical screws which move the matrices are quite visible (there are three of them, two in front of the matrix and one behind it - since this is a view from the back of the machine, "in front" of the matrix means, in this view, to the right), but the distributor bar is just a sort of black line almost hidden between them. Several matrices are on the distributor bar, while two others have fallen off and are on their way to the magazine (through an invisible channel entrance, thanks to the miracle of 1940s vintage image processing).

(From The Big Scheme of Simple Operation. (Brooklyn, NY: Mergenthaler Linotype Company, 1940). Page 33.)

Moving around to the front of the machine again, the matrices slide down into the channels of the magazine as shown below, and await release from the bottom of the magazine by the keyboard/keyframe/escapement mechanism.

(From The Big Scheme of Simple Operation. (Brooklyn, NY: Mergenthaler Linotype Company, 1940). Page 8.)

For a partial reality check, here are images of actual matrices for "e t a o i n":

(These are straight out of a Linotype/Intertype Model X 90-Channel magazine as it ran at a small newspaper in rural Wisconsin. If you've never seen a Linotype mat before, I should perhaps note the obvious: that the white printing on them in the first image is something I did in the picture; it isn't on the mats themselves. The less vivid printing on the "narrow" side of the matrix, in the second image (the "reference" side), is what the operator would see. The distributor, of course, "sees" only the teeth. The study of the notches in this font of mats is a lesson in keeping a small newspaper running in adverse conditions, but that's a topic for a different Notebook, as is the identification information on the other wide side of the mats (not shown.))

(In case you're interested, these are mats from a font of Linotype 9 triangle 754, which is 9-point Helvetica with Italic. When the Linotype which was running this font was built (1912), one of the designers of what became known as Helvetica (Max Miedinger) was two years old.)

Below, I'll try to bring this together for an entire 90-channel magazine, in a single diagram using only ASCII characters (which should be cut-and-pastable anywhere, without losing formatting). In the text/diagram below, the first two lines are the tooth combination interpreted as a binary number (reported in decimal). Read the numbers vertically: 00, 01, 02, 03, ... Similarly, two lines of numbers near the bottom represent channel numbers in a "90-channel" magazine. These start out with "q" (for "quad tray"), which isn't in the magazine, and "x" (for not-used), which also isn't in the magazine. At the very bottom, I've shown some of the characters commonly (but not always) associated with this channel (there are variations not shown here, and I've used "." to represent untypable characters, but this should give a general sense of place relative, at least, to etaoin shrdlu). In the middle are the tooth combinations, with "T" repesenting the presence of a tooth, and "." representing its absense. Note how the pattern is pure binary counting, starting at 2.

0000000000111111 1111222222222233 3333333344444444 4455555555556666 tooth combo 
0123456789012345 6789012345678901 2345678901234567 8901234567890123 as a number 

xx.T.T.T.T.T.T.T .T.T.T.T.T.T.T.T .T.T.T.T.T.T.T.T .T.T.T.T.T.T.T.T 
xxTT..TT..TT..TT ..TT..TT..TT..TT ..TT..TT..TT..TT ..TT..TT..TT..TT 
xx..TTTT....TTTT ....TTTT....TTTT ....TTTT....TTTT ....TTTT....TTTT 
xx......TTTTTTTT ........TTTTTTTT ........TTTTTTTT ........TTTTTTTT 
xx.............. TTTTTTTTTTTTTTTT ................ TTTTTTTTTTTTTTTT 
xx.............. ................ TTTTTTTTTTTTTTTT TTTTTTTTTTTTTTTT 
xx.............. ................ ................ ................ 

qx00000000001111 1111112222222222 3333333333444444 4444555555555566 90-channel 
qx01234567890123 4567890123456789 0123456789012345 6789012345678901 magazine 
..eetaoinshrdluc mfwypvbgkqjxz... ...,.:;?.(|`!-.) .'*1234567890$.E 

6666667777777777 8888888888999 tooth combo 
4567890123456789 0123456789012 as a number 

.T.T.T.T.T.T.T.T .T.T.T.T.T.T. 
..TT..TT..TT..TT ..TT..TT..TT. 
....TTTT....TTTT ....TTTT....T 
........TTTTTTTT ........TTTTT 
................ TTTTTTTTTTTTT 
................ ............. 

6666666677777777 7788888888889 90-channel 
2345678901234567 8901234567890 magazine 

4. The 90-Channel Keyboard (Standard)

The pattern of teeth, their dis-engaging locations on the distributor bar, and the channels into which they drop all fit together nicely in the case of the 90-channel magazine, and fit together also with the standard "90-channel" Linotype keyboard:

(From Useful Matrix Information. (Brooklyn, NY: Mergenthaler Linotype Company, 1937). p. 84.)

This layout shows the standard keyboard arrangement for general work using a 90-channel magazine. "Fractions pi" means that fraction logotypes (e.g., "3/8" as a single character) would be supplied with all seven teeth and would therefore run through the entire distributor without going into the magazine; they'd emerge on the far side of the distributor and be routed to the "pi stacker" on the operator's right side.

Read the layout from top-to-bottom, left-to-right: "e t a o i n" then "s h r d l u", and so forth. The numbers underneath the keys indicate the magazine channel to which they key is connected. Whatever matrix is in that channel - however it got there - is what drops into the assembled line, regardless of what might be on its casting or reference face.

In case you should happen not to be familiar with the Linotype keyboard (and if this is the case, perhaps there are other things you should be reading before this rather detail-level study), here is a picture of one:

Here's a closer view of the keyboard; the nonalphabetic characters are a bit fuzzy due to the halftone original, but they general "etaoin shrdlu" layout should be clear.

(From The Big Scheme of Simple Operation. (Brooklyn, NY: Mergenthaler Linotype Company, 1940). Page 13. This booklet appeared also as a section in several other Linotype publications.)

Finally, here's a cut-away view of how it all fits together: a key being pressed (lowercase 'v' on the keyboard) actuates a series of levers, cams, and rods which causes a matrix (item 18 in the diagram) to be released from a channel (channel no. 19, though the digram doesn't say this) in the magazine (item 17 in the diagram). There is much more to the keyboard mechanism than this, of course; I'm only including this here for general orientation.

(From Linotype Machine Principles. (Brooklyn, NY: Mergenthaler Linotype Company, 1940). Page 48.)

5. The 72-Channel Magazines

In the very earliest Linotypes from the Model 1 (the "Simplex" Linotype) of 1890 through about the Model 4, there were various changes in magazines that I don't yet understand, and so cannot comment upon. By (or with?) the introduction of the Model 5 in 1906, however, the current "90-channel" magazine became standard. (Indeed, many such magazines are called "No. 5" magazines precisely because they were of the type first associated with the Model 5.)

Any particular magazine configuration will have built into it limits as to the width of the individual channels (which in the 90-channel magazine vary in width depending upon the size of the character designed to run in the channel; it would be wasteful to make the channels intended for 'i' as wide as those intended for 'M'). This means that, because type gets wider as it gets taller, above a certain point size (depending on typeface) matrices which might otherwise work on the machine (in the casting unit) won't fit in the magazine. [TO CHECK: what size is this? is it 14-pt max for the 90-channel? Does it vary with some typefaces; condensed?]

If it isn't possible to accomodate this by making the magazine wider, then one other possibility is to accomodate it by having fewer, but sometimes wider, channels. I am presuming (but have not seen verified - so I'm just guessing) that this is the origin of the 72-channel magazine.

The 72-channel magazines fit in the same space, more or less, as the 90-channel magazines. They also cover the same range of numeric tooth combinations (2 through 92), but there are some interesting exceptions to this.

Given that the standard character set for a 72-channel arrangement is a proper subset of the standard 90-channel arrangement, my own intuition - had I been a Linotype engineer in the early 20th century - would be to keep the same characters at the same tooth combinations but simply to omit those not used. This would have preserved (with gaps) the binary counting sequence. This is almost, but not quite, what was in fact done. The "not quite" is a puzzling exercise in history, I think.

Here, to start with, is a pure 72-channel keyboard, laid out with the intent that it should run a font designed for a 72-channel magazine:

(From Useful Matrix Information. (Brooklyn, NY: Mergenthaler Linotype Company, 1937). p. 100.)

This seems straightforward. It's just like the 90-channel keyboard, but with 72 keys. The numbering (remember: these are the channel numbers) is sequential and runs in a similar order, top-to-bottom, left-to-right. It would appear to be very simple.

A glance at the chart of tooth combinations sorted by the combination (not by the channel or key) indicates, however, that it is not so simple:

linotype-tooth-combination-comparisons.pdf or linotype-tooth-combination-comparisons.ods

The order works just fine through the lowercase alphabet; it's identical to the 90-channel organization. Lowercase 'z', then, has a binary tooth pattern of 28 (0011100, or 3-4-5) and runs in channel 26.

What happens next is best expressed in a short table:

Binary Pattern Teeth Present Intended Character Channel (72-Channel Magazine)
28 0011100 3-4-5 z 26
29 1011100 1-3-4-5 Em Space 30
35 1100010 1-2-6 Comma 27
36 0010010 3-6 Period 28
38 0110010 2-3-6 Semicolon 29
40 0001010 4-6 Fig. Space 36
46 0111010 2-3-4-6 Thin Space 42
51 1100110 1-2-5-6 1 31

From here, things go back to "normal." The numerals do in fact rise in ascending binary order from 1 (tooth combination 51 (binary) to 9 (59 binary) and 0 (60 binary). Of course, since channel 36 is already used, the channel numbers for the digits are not sequential - but the tooth count numbers are. There's a short skip after this, and then "ETAOIN..." picks up as normal with the capitals.

Significantly, in both the 90-channel and the 72-channel layouts the binary tooth patterns for lowercase ("etaoin..."), uppercase ("ETAOIN...") and the numerals ("1234..0") are identical. The channel numbers may differ, but the machine doesn't really "know" about channel numbers. It knows only about tooth combinations, and they're the same. This means that the alphanumeric parts of a 72-channel font have full binary compatibility with a 90-channel magazine (how's that for good computer engineering decades before the first computer!)

But what happened in the middle?

I've never actually seen a 72-channel Distributor Bar (I'm trying to remedy this), and there are no sufficient pictures in the Linotype or Intertype literature (I've checked pretty thoroughly, though I have not yet checked the patent literature). This is therefore just "reverse engineering" on my part. But each position on the distributor bar must correspond to the channel number into which it drops matrices. Thus, on a 90-channel bar the positions are numbered 0-90. These correspond sequentially to binary number patterns 2-92. On a 72-channel bar, the positions must be 0-72, but if the tooth-to-channel correspondences are as they are, then the tooth pattern cannot be a sequential pattern of binary numbers, or even a numerically increasing pattern with gaps.

Channel 26 corresponds to binary pattern 28, and channel 27 corresponds to binary pattern 35, which is greater. This merely skips. But then when channel 30 comes up later, its rail pattern is a binary 29 - one of the numbers that was skipped earlier.

Slightly later, channels 36 and 42 look at first to be ok but when the numeral start up in channel 31 it becomes clear that they came too early.

If this is a little confusing (and indeed it is; it makes my head spin), look back at the "72-Channel Face" keyboard diagram, above. Start with 'z' in channel 26. Going ahead from 'z', the next characters should be Comma, Period, and Semicolon. By channel number, they are indeed the next characters (channels 27, 28, and 29). On the Distributor Bar, they should correspond to increasing binary numbers (possibly skipping some, but always increasing). They do: 35 (binary), 36, and 38. But then the next key, Em Space, in the next channel, 30, has a binary tooth pattern that has already been skipped, 29.

The key series "5 EnSpace 6" ("EnSpace" is the same as "Figure Space") uses channels 35, 36, and 37, but these correspond to the binary distributor bar patterns "55 40 56". Later, the key series "0 ThinSpace E" (channels 41 42 43) corresponds to binary distributor bar patterns "60 46 63".

I really don't know, for certain, the answer to this. (This really is research in progress.) But I'm willing to hazard a guess that it has to do with getting a single Linotype to use both 72-channel and 90-channel magazines. Here's a standard keyboard arrangement for such a combination: a 90-key keyboard, set up for a 72-channel face in a "72-90" or so-called "2in-1" Linotype:

(From Useful Matrix Information. (Brooklyn, NY: Mergenthaler Linotype Company, 1937). p. 98.)

I didn't mention it as such earlier, but it should have been relatively obvious that the only "out of place" characters in the 72-channel situation were the three space characters (Em, En/Fig, Thin). In the pure 72-channel keyboard, they were by themselves down at the bottom of the keyboard (and received their channel locations from that: channels 30, 36, and 42, which are logical for this arrangement). When using a 90-key keyboard, however, these three spaces are "back where they should be" - that is, they're in their standard location for a 90-channel keyboard.

This explains En/Fig Space and Thin Space, I think. Their tooth combinations are identical in both 90-channel and 72-channel arrangements (En/Fig Space is binary 40 = 0001010 = 4-6. Thin Space is binary 46 = 0111010 = 2-3-4-6.) Making these the same tooth pattern means that 90-channel-font En/Fig and Thin spaces can run in 72-channel fonts. The only disadvantages are that their distributor bars don't have the most logical pattern (not really a disadvantage; you can make it anything that works) and there must be some mechanism in a 72-90 machine to let these keys, which are to the left of the keys for channels 31-onward, operate channels to their right.

This almost explains EmSpace. It's in the "right" location for a 90-channel keyboard, but actually the binary tooth combination for a 72-channel EmSpace (29 = 1011100 = 1-3-4-5) is not the same as that for a 90-channel EmSpace (32 = 01000010 = 1-6). However, there is some precedent for running an EmSpace in 90-channel magazine channel 30 (normally the little-used ligature "ffi"). Here's a keyboard layout with such a location for Em Space. Perhaps not insignificantly, it's diagram number, 6, makes it the lowest-numbered (and thus oldest? I'm not sure) keyboard diagram in the 1937 edition of Useful Matrix Information.

(From Useful Matrix Information. (Brooklyn, NY: Mergenthaler Linotype Company, 1937). p. 88.)


In other words, it all makes relatively good sense if you work it out logically and historically, in sufficient detail. The Linotype engineers were doing the same thing - responding logically to then-current needs - all those decades ago.

6. An Aside on Computers

The dominant feature of the computer industry is not, as it might have you believe, innovation; rather, it is compatibility. Once a computer system is established, compatibility with it is far more important than any actual virtues it might or might not possess.

Technically, software compatibility comes in two forms: "source compatibility" and "binary compatibility." Computer programs are written in laguages that people (mostly) understand. These are called "source" programs. They're translated (by other programs) into versions that computers can execute. These versions are commonly called "binaries," because although they actually consist of computer commands and data, to a human unfortunate enough to see them they look like large masses of binary numbers.

From one computer system to its successor, source program compatibility is hard, but "binary compatibility" is even harder. If you can still take the exact same program you bought fifteen years ago and plug it into your brand new computer and run it, then you have binary backward compatibility. (Chances are you don't; try it someday.)

From a computer programmer's perspective, what Ottmar Mergenthaler achieved with the 90-channel magazine/distributor was a brilliant and probably isolated re-invention of binary counting. What the subsequent engineers at the Mergenthaler Linotype Company achieved with the 72-channel magazine/distributor, and with the integration of the two in 72-90 channel machines, was binary backward compatiblity - one of the holy grails of commercial computing - long before the first computer appeared.

Very impressive. Very cool.

7. The 72-90 Mechanism

The use of out-of-sequence binary patterns for 72-channel magazines, by itself, requires only special distributor bars for 72-channel distributors. These would be required in any case; all that matters is to ensure that the changes in the pattern don't allow the matrices to fall off the bar too soon.

The placement of the three sapce keys on the 72-90 keyboard arrangement, however, requires a special mechanism in the keyboard and keyframe. Some changes would be necessary in any case, of course, since the channels in a 72-channel magazine don't line up with those of a 90-channel magazines. There were at least two such mechanisms for the Linotype, and Intertype had its own solution (or solutions?) I have not yet investigated any of them.

8. The Side Magazines

There are some peculiarities to the tooth patterns used in matrices for the side magazines. I'll investigate these soon, I hope.

About the images

Select Resolution: 0 [other resolutions temporarily disabled due to lack of disk space]