Post by Chris HofstaedtlerPS: I never understood why there's both w and who, and why they are
different implementations.
As I understand it, it's the same reason as why more(1) vs. pg(1) and
why lp(1) vs. lpd(1)--the AT&T/USG/USL vs. BSD schism.
who(1) goes all the way back to "First Edition" Unix (1971).
https://minnie.tuhs.org/cgi-bin/utree.pl?file=V1/man/man1/who.1
w(1) came along much later, in 3BSD (late 1979).
https://minnie.tuhs.org/cgi-bin/utree.pl?file=3BSD/usr/man/man1/w.1
About that time, the BSD/USG Unix split occurred. AT&T's Unix System
III (1980) did not incorporate the Berkeley CSRG's w(1); instead they
aped it with a now-forgotten administrative utility called whodo(1).
https://minnie.tuhs.org/cgi-bin/utree.pl?file=SysIII/usr/src/man/man1/whodo.1m
System III became System V (1982), which continued to eschew w(1) along
with numerous other Berkeley CSRG innovations for much of the duration
of its commercial relevance. In places where ecumenicism was more
important, as with Research Unix (the AT&T CSRC at Murray Hill, New
Jersey, where Unix was born), some "vendor" Unices like Sequent's DYNIX
(with its two "universes" either of which any given user [or process?]
could be assigned), and eventually Unix System V Release 4.2 (1992)--I
think--and Solaris, the two utilities came to coexist. Supporting both
was a Simple Matter of Programming, especially once kernel development
settled down.
...and, as I understand it, the kernel was why there were different
utilities in the first place.
AT&T USG/USL and Berkeley had divergent concerns and development
priorities in the Unix kernel itself. Back in those days, utilities
like ps(1)--and I think who(1) and w(1) as well--had to be setuid root
so that they could read kernel memory. This was in the days before
/proc. It took a while for the idea of a pseudo-file system to expose
kernel memory structures to user space in a stable and safe way to come
to fruition.
Since there was no API for obtaining process or user session
information, the different development organizations, with their
divergent kernels, changed what ever they needed to in the kernel to
achieve other objectives, like implementing virtual memory with demand
paging, or supporting symmetric multiprocessing.
Regards,
Branden