[mjn]

I think the other factors in play are the dominant ones in the Unix/Plan9 issue, though. The original "upstream" of Unix was Research Unix, the stuff the Bell Labs researchers were working on. V1 came out in 1971. But V7, released in 1979, basically got forked into a bunch of commercial Unixes (and indirectly, some free ones via the BSDs). Research Unix V8, V9, and V10 got virtually no uptake, despite being incremental evolutions of the original upstream, due to a mixture of inertia and licensing restrictions from AT&T. Instead, different incremental evolutions of V7, developed by other developers outside Bell Labs, became the dominant strains of Unix.

Plan9 was a pretty significant departure from Research Unix V7, but not as significant a change compared to V10... which didn't matter because nobody outside Bell Labs was using V8/9/10 anyway, since everyone was working on various forks of V7 (BSD, SunOS/Solaris, Xenix, etc.).

[4ad]

I agree Plan 9 is not a good example in this story. Plan 9 didn't failed because of the problems associated with a re-write as described in Joel's post, but because of political and social issues.

> Research Unix V8, V9, and V10 got virtually no uptake, despite being incremental evolutions of the original upstream, due to a mixture of inertia and licensing restrictions from AT&T.

I'm not sure what you mean by inertia. v8-v10 were not widely released to avoid another BSD incident.

v8-v10 were 4.3BSD derivatives, so I'm not sure I'd call them incremental evolutions on the original upstream. They removed and/or reworked all the new BSD stuff though, and the basic userland was very custom. Some of it survives (in some form) in Plan 9.