Casper.Dik at Sun.COM wrote:
>> First off, I really like what this case is trying to do. But I do have
>> a possible concern: /usr/ucb/ps could have been used with a leading
>> "-". E.g. /usr/ucb/ps -aux and /usr/ucb/ps aux both return the same thing.
>>
>> I'd humbly suggest that if getexecname returns /usr/ucb/ps then the
>> legacy UCB behavior should be used unconditionally.
>>
>
> What I propose isn't all that different. /usr/ucb/ps will ONLY behave
> like /usr/bin/ps IF AND ONLY IF one of the arguments is NOT recognized by
> /usr/ucb/ps. Rather than printing an error message, it tries to do
> something plausible.
>
> I thought the examples clarified that.
>
Hmm... maybe I didn't understand. As long as /usr/ucb/ps behaves as
/usr/ucb/ps whenever any valid syntax that was accepted by it today is
given, then I'm ok with it. (And understanding that "-" is a valid part
of the syntax. :-)
That said, I do sort of think that the rule you have, while workable, is
more confusing and violates the principle of least surprise. (I.e.
you added a switch by accident which suddenly changed the meaning of
other options in a totally unexpected fashion.)
But given as we're talking about a case that would otherwise be an
error, I'm not sure there is architecturally relevant cause to debate it
further.
-- Garrett
