Documented Unix facilities (across 93 major Unix releases)

https://dspinellis.github.io/unix-history-man/

by adg001 • 4 years ago

30 5 4 years ago

dspinellis.github.io

What a great surprise (but not shocking) to see it's a work of the author of "Code Reading: The Open Source Perspective" [1].

[1] https://www.spinellis.gr/codereading/

asveikau2• 4 years ago

When they list first appearance, it's weird that the only modern system they consider is FreeBSD.

There are a few interfaces that were introduced in Linux, OpenBSD, NetBSD etc. that saw subsequent broad adoption across all of modern Unix-like OSs, FreeBSD included.

Maursault• 4 years ago

It is very common to list Linux among the UNIXes, and Linux is similar, will operate the same with nearly all of the higher layers and applications, but it is different enough underneath and fundamentally so that while a UNIX administrator can rapidly and fluidly go from NetBSD to OpenSolaris and Illumos to AIX to FreeBSD and OpenBSD without much pain, they will experience pain when first exposed to Linux, and similarly, a Linux administrator or developer that is only well-experienced in RHEL, CentOS, Fedora, SUSE and/or Debian will experience pain when first exposed to AIX or Solarises or BSDs.

For whatever reason, Linus or the distributions themselves arbitrarily deviated from the ways common in all UNIX, in directory structure and tools and other details, that in one sense they are still similar, brothers even, but in another they are entirely different, and distant cousins removed several times, and, again, I believe these deviations were entirely arbitrary, and as such, unnecessary.

/rant, nothing personal, everyone considers Linux as UNIX, and they're not entirely wrong, but the differences are neither minor nor pedantic, and I, for one, wish it had been different and Linus or GNU or distros had not made these unnecessary deviations from standard UNIX and continued to do so, as the advantages of the deviations, if there are any, are negligible. To me they seem different for the sake of being different, unnecessarily adding complexity to an already complex matter.

And I acknowledge that Linus may be entirely innocent and the problem may have always been with GNU, I'm just not certain who the culprits are, but they seem to have made the mistake of changing the architecture without being aware of what came before, falling into the tragedy of reinventing wheels.

asveikau2• 4 years ago

I'm not talking about whether Linux is Unix. I'm talking about the indisputable influence it has on other Unix-like OSs in the modern era. When talking about where a syscall was first introduced, for historical accuracy, Linux should be correctly cited when appropriate.

I also said OpenBSD and NetBSD, but this seemed to escape your multi paragraph rant.

Also worth noting that FreeBSD, while technically derived from Unix in a way that Linux cannot claim, still today cannot use the Unix trademark.

jlokier• 4 years ago

Some interesting anomalies that look like errors in the tables. E.g. setjmp() present in 32V, BSD3 and BSD 4.2, but longjmp() not present. That doesn't make sense and doesn't match their man pages.

Documented Unix facilities (across 93 major Unix releases)

Related Stories