Hacker News new | past | comments | ask | show | jobs | submit login
The Paige Compositor: First US Typography Machine (asme.org)
10 points by Bluestein 3 months ago | hide | past | favorite | 4 comments



It's nice to see the picture. I knew that Clemens lost money on it, but not what it looked like.

It's a mechanization of manual typesetting. You have a huge stock of type, which is fed through the machine and assembled into lines. Those are the actual type, which then goes into a form to make up a page. This ties up the type supply for hours or days. After printing the pages, all that type has to be run through a sort to get it back into the right slots. There's a video of the machine here [1], but the machine no longer works, and it's not clear how the sorting process worked. The ASME page says it had automatic "distribution", but how?

Linotypes use brass matrices to assemble a row of moulds, which are then filled with lead to make a single piece with one line of type.[1] The matrices have their identity binary-coded into slots at the top. They're sorted by a simple and clever mechanism where they're pushed along a bar with notches that match the coding of the matrices. The matrices fall off the bar when the notch patterns match, which puts them back in the proper channel for that letter. (It's not a binary match. It's a 7-bit pattern, which, if when ANDed with the distributor bar pattern, yields all zeroes, causes the matrix to fall off. Here's the coding.[2] 7 bits, but only 90 valid patterns.)

Linotype matrices are much taller than one line of type, allowing for plenty of room for the coded slots. A machine which actually sets type has to set type slugs no taller than the line height, so the Linotype's simple sorting system won't work. I wonder how Paige did it. The picture clearly shows the typesetting part, and there's something off at the left that must have something to do with type distribution. Anyone know how this worked?

The claim is that the Paige machine was 60% faster than the Linotype. But that was probably true only for early Linotype models. In later Linotype models, the line caster is double-buffered, with one line of matrices being assembled and one being cast. Distribution is concurrent with typing and casting. Videos of Linotypes in use show the user typing fast and not waiting for the machine.

There were other early typesetting machines. All worse.[3]

[1] https://archive.org/details/0066_Typesetting_Linotype_02_25_...

[2] https://archive.org/details/MLCUsefulMatrixInformation_610_1...

[3] https://metaltype.co.uk/wpress/early-typesetting-machines/


I'm late, but just came by to say thanks for the comment.


Thanks for the amazingly informative comment.-


As I recall, Mark Twain lost the equivalent of today's $9M and went bankrupt as a result.-




Consider applying for YC's W25 batch! Applications are open till Nov 12.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: