"auto" for Kerberos, a history

Douglas Bagnall douglas.bagnall at catalyst.net.nz
Thu Aug 20 23:36:24 UTC 2020


On 20/08/20 6:53 pm, Stefan Metzmacher via samba-technical wrote:
> 
> yes means no fallback to NTLM,
> 
> Should we use "disabled", "if_available", "required"
> instead of "no", "auto", "yes"?
> 

I am a fan of "auto", as telling the tool "you decide".

It is nice for the user because there is a convention, and it is nice for
the toolmaker because there is no implicit guarantee about what decision
it will make.

There is considerable precedence for --foo=auto in command line culture.

The biggest cluster is --color=WHEN, where WHEN is conventionally
always|auto|never. This is shared by  ls, rustc, rpm, tc, grep, ip, git,
and dozens or hundreds or thousands of other tools.

While --color=WHEN's always/never is a different take than yes/no, yes/no
are usually also accepted:

  $ ls --color=maybe
  ls: invalid argument ‘maybe’ for ‘--color’
  Valid arguments are:
    - ‘always’, ‘yes’, ‘force’
    - ‘never’, ‘no’, ‘none’
    - ‘auto’, ‘tty’, ‘if-tty’
  Try 'ls --help' for more information.


A very brief search for other --foo=auto options reveals:

cp --reflink=auto --sparse=auto
xz --format=auto
flake8 -j auto
git log --decorate=auto
git format-patch --cover-letter=auto --numbered=auto --base=auto
etags --language=auto
gpg --trust-model=auto --tofu-default-policy=auto
valgrind --expensive-definedness-checks=auto
crash --kaslr=auto
grub-install --core-compress=auto
mount -t auto


Now, of course, I am not saying we want to follow gpg's human interface
guidelines, just that this pattern will have some resonance with users. If
the option is "if_available", people like me will be always forgetting the
exact term and whether it uses an underscore or a hyphen.

OK, so sorry for being late and somewhat bikesheddy. I don't object to new
more precise terms, but please let's not remove "auto". It is alright to
be vague, because it is SUPPOSED to be vague.

cheers,
Douglas



More information about the samba-technical mailing list