Re: inconsistent field splitting

[Pierre Gaston]

On Wed, Aug 18, 2010 at 12:36 PM, Marc Herbert <Marc.Herbert@gmail.com> wrote:
> Compare:
>
>   for a in "$(echo 1 2)"; do echo "x${a}x"; done
> x1 2x
>   for a in  $(echo 1 2) ; do echo "x${a}x"; done
> x1x
> x2x
>
>   a="$(echo 1 2)"; echo "x${a}x"
> x1 2x
>    a=$(echo 1 2);  echo "x${a}x"
> x1 2x
>
>
>
> Shell quoting is difficult enough; why is such an inconsistency making
> it even more confusing?
>
> Uwe Waldmann might give a clue in his excellent "Guide to Unix shell
> quoting":
>
>   Note that in these [assignment + others] cases, the shell syntax
>   allows only a single word, not a sequence of words, so that blank
>   interpretation or expansion of globbing characters might result in
>   something syntactically illegal.
>
> In other words:
>
>   In order to save you from some very obvious syntax errors, I'll
>   make quoting even more confusing than it already is.
>
> Sorry but I am not grateful at all.
>
> Or is there a better rationale for this design?
>

What should a=* or a=$(echo 1 2) do?
Assign only the first file or just 1 to a? or result in an error?

This hardly seem more consistent  or less surprising to me.

There are other places where word splitting is disabled, in case,
inside [[ ]]...etc...