20 Years Ago in UNIX

by Peter H. Salus
<peter@eng.us.uu.net>

USENIX Historian

If you have your System III or PWB documentation handy, you might look at the title page. You'll find Ted Dolotta's name prominently displayed.

Ted was at Princeton, Bell Labs, and INTERACTIVE Systems in the '70s and '80s, and he just retired from Softbank, where he was vice president. He went to INTERACTIVE after a genuine courtship:

One of the lures that Peter Weiner (who was a colleague of mine when we were both on the Princeton faculty) employed to lure me to INTERACTIVE was to give me his California UNIX license plates, which I recently sold via auction to John Mashey for $6,000, giving the money to the John Lions Scholarship Fund (and the charitable tax deduction to John Mashey).

Ted is a reader of these articles, and after my query about System IV he sent me the following:

Hi, Peter,

I noted your query in ;login: regarding System IV UNIX at Bell Labs.

As best I remember, there was indeed a System IV, but it was never offered for licensing via Western Electric. I no longer recall the exact reason why, but I think that it was not a major step beyond System III.

One of the major improvements was a full set of application manuals in two volumes, some 1,000 pages (sed, troff, yacc, etc., etc. — some 50 or more manuals) which I edited for consistency and coherence and cross-references, etc. This document set was distributed internally with System IV, and externally with System V license. I left Bell Labs to join INTERACTIVE Systems right after this 2-volume set went to the printer, but before it was actually distributed.

Also, System V was already in the planning & early development stages, so someone (Western Electric?) must have opted to wait for System V for licensing purposes and skipped licensing System IV.

Things lead to one another. Discussing the System III documentation with Ted, I elicited the following:

Until we started doing System III at Bell Labs, all UNIX manuals were standard, letter-size (8.5" x 11"), loose-leaf manuals, and every update had only *new* pages and an update sheet: "Remove pages . . . , insert pages . . . ").

When we were doing System III, I got to thinking: we were about to send out N thousand update packages to N thousand users within the Bell Systems (never mind licensees outside the Bell System), and we'd be propagating the fiction that *every one* of these highly-paid individuals would take the time to punctiliously update his/her copy of the manual. Right. We knew that, for practical purposes, every copy of the manual was unique — no two were alike. A genuine support nightmare. . . . And in terms of wasted technical talent hours spent updating thousands of manuals, mind-bogglingly expensive.

So I decided to do something better: I observed that Bell Labs gave every one of its 20 or 30 thousand employees a brand new, bound company phone book every six months, with white pages (alphabetical) and yellow pages (by department) . . . the whole nine yards. And it wasn't just for expensive techies — it was for everyone. . . . So why not do the same with UNIX manuals, I asked myself. My management gave me the expected answer: we had no money to thusly "subsidize" all users of UNIX who were not in our department.

I appealed my case, and finally played the BTL phone book trump card, and, to my surprise, I won. I was authorized henceforth to issue a brand new, bound manual with each major release.

Using the BTL phone book as a model, I realized that I could not use the "perfect" binding — the manual would never stay open to any one page — so I decided to use a comb binding; but I liked the 6" x 9" size of the BTL phone book and decided to adopt it for the UNIX System III manual.

I modified the -man troff macros by adding the -rs1 option ("s" for "small") and re-typeset the entire manual. To my amazement, only a dozen or so pages "broke" and required fixes (and in most cases, it was just careless initial coding that simply needed cleaning up).

The new format was a big success, and, although I can't prove it, I believe it was the first widely distributed computer manual in that size. Then a while later, IBM came out with the PC and used the 6" x 9" format, and the rest, as they say . . . If my belief is true, then most paper computer manuals today owe their size to the Bell Labs phone book. It's not as good a story as the one that derives the size of the space shuttle boosters to the size of the hind quarters of a Roman war horse, but it's my very own.

I love these stories, among other things because they frequently demonstrate just how clueless large corporate entities are. Here's a final Ted Dolotta anecdote for this article.

Speaking of loose-leaf binders, you may be aware that before its breakup in 1983, the entire Bell System was standardized on 8.5" x 11" FOUR-ring binders — two at the top, two at the bottom. But because the rest of the world used 3-ring binders, to help communications with the outside world Bell System punched paper had 7 holes, so it would fit into either 4-ring or 3-ring binders. This worked OK until a newly minted manager on a rotational assignment at Western Electric (which then purchased or manufactured everything for the Bell System, from paper clips to Central Offices) noticed that all stockrooms had 7-hole paper but 4-ring binders. He fixed the situation by decreeing that henceforth all binders would be 7-ring. And it was done. And when Western Electric purchased things, it purchased them in multi-railroad-car lots.

Thanks, Ted.