Nicholas Bamber dixit:
> Please could Thorsten and Dominik please work out who will be responsible for
> mksh in Debian? Can you work together if that makes sense? I can sign and
> upload if necessary.
OK, we had a short talk on the office floor about this ;-) and
I’ll be putting myself back into the Maintainer field and take
formal responsibility for this as external contributor.
> 2.) ksh93 and mksh are alternative implementations of ksh (with subtle
> differences that should be documented). In this case both can coexist on
> Debian
> but one (potentially either but by default ksh93) should assume the role of
> ksh.
This is what is currently true, and the desired outcome:
• scripts specific to ksh93 use ksh93 in the shebang
• scripts specific to mksh use mksh in the shebang
• scripts using ksh in the shebang can be…
– old pdksh scripts,
– old ksh88 scripts,
– or scripts written for “any ksh subset” – I did a short
UDD search just now, and all cases of this in Debian
also accept zsh as ksh providing package.
This means that, if only one of ksh93 or mksh is installed,
that package takes over ksh, but if both are installed, the
official Korn Shell should take over ksh (unless the local
admin overrides, of course).
All use cases in Debian use alternative package relationships,
i.e. don’t force the installation of one specific variant but
don’t use a virtual package either. Let’s not introduce one.
Renaming ksh93 from ksh to ksh93 would also make backporting
much harder. KISS.
bye,
//mirabilos
--
I believe no one can invent an algorithm. One just happens to hit upon it
when God enlightens him. Or only God invents algorithms, we merely copy them.
If you don't believe in God, just consider God as Nature if you won't deny
existence. -- Coywolf Qi Hunt
