> AutoCAD is a transformative product

It really was. It replaced manual drafting. Drawing revision before AutoCAD involved maintaining a master drawing and updating it manually with pencils and erasers. Final drawings were inked in. Copies were made by running master drawings through a blueprint machine. When there were too many revisions, someone had to redraw the drawing by hand, and the copy had to be checked by hand by a checker.

There were CAD systems before AutoCAD, but they either required a more expensive computer than an PC, or they couldn't handle a drawing too big for memory. The big innovation in AutoCAD is that it had a paging system for working on drawings too big for the machine. The code was paged in and out in segments. The drawing was paged in and out in sections.

(I did some of the early AutoCAD ports to non-IBM PCs. Compatibility hadn't been established yet. Everything needed a driver. AutoCAD had "more drivers than Yellow Cab" at one point.)

Interesting. As I recall, we never thought of the paging as a big innovation: just what had to be done in order to do anything in a memory space of 128K bytes. You can bet that it presented an interesting challenge in adding the UNDO feature to the existing system. BTW it seems to me that AutoCAD never got much recognition for its technical innovations. One very early example: It was almost certainly the very first program which could run equally well with or without the 8087 floating-point chip (ahem - except for speed!), needing no special user configuration or anything. Credit where credit is due: Not a Walker idea, but an inspiration of one of the unsung super-hackers among the original crew.

On balance, CAD has improved the industry for sure. But, because I'm old enough to yell at clouds, I would like to point out the flip side.

In manual drafting days, changes were hard, as you note. But because of that, there was a LOT more pre-planning, because once you start putting the Koh-I-Noor to mylar, changes were difficult. So you avoided them whenever possible. Architects had to sit down with owners and say, "look, past this point, you don't get to make changes, at least not for free."

Now, buildings are quite often designed-by-addenda. The due date is just a date. You get extra time to actually "finish" the job by issuing massive addendums prior to bid. And because of that, architects don't make owners sit down and tell them everything, so you'll get A/V or finishes or whatever really, really late in the design phase. "No problem, we'll fix it by addendum."

I am also a bit sad that drafting skills have died out somewhat. I've seen some really, really beautiful bluelines. I've heard of electrical engineers who would make smiley-faces with their homerun arrows. And the process of manual lettering is very zen, and teaches people spatial awareness like nothing else.

All that said, man alive, I love AutoCAD. Embedding Lisp in the program was amazing. The modern thing is now Revit, which has some good things going for it as well, but it is not (and may never be) anywhere near ACAD for a lot of work. RIP John.

AutoLISP was put in because it was a memory-safe interpreter. Allowing users to extend AutoCAD with C would have created a debugging nightmare. AutoLISP could detect its own errors without crashing AutoCAD. LISP was the only game in town back then for interpreters which could deal with variable-sized data.

AutoCAD users were not programmers. Computer knowledge was not widespread. Keeping users from losing their drawing files, a major long-term asset, was crucial to product acceptance. There was a strong "don't screw up" ethos within Autodesk. Much of that came from the founders, who were mostly mainframe operating system programmers.