Re: Subcommand dispatcher and alias factory

[Daniel Shahaf <d.s@daniel.shahaf.name>]

Bart Schaefer wrote on Fri, Aug 31, 2018 at 16:01:25 -0700:
> the shell MUST first execute "git".  Unless all commands with
> subcommands were to share some standard exit status or error message
> that would allow the shell to discover that the reason for failure was
> that the subcommand is not found, the shell has no generic way to know
> why the primary command failed in order to try it again differently.
> You can program it separately for each command that has subcommands,
> as was shown in examples in a couple of the links you mentioned.
> 

I don't understand what's so bad about subcommands.  What's different
between, say, «git ci -S foo» failing because 'git' doesn't know the
'ci' subcommand, and that command failing for any other business logic
reason (e.g., 'foo' being an invalid argument)?

> This, by the way, is why commands with subcommands are basically
> horrible, and a perversion of the UNIX command model.  It's like
> creating a custom shell within the shell.  Which latter is fine if you
> tell the user that's what you're doing (I was one of the major
> contributors to a custom shell for managing email, back in the day),
> but in the git-style incarnation always grates on me in the same way
> that layering violations in object-oriented programs do.