Casper.Dik at sun.com wrote:

>> However I haven't seen any justification as to why the ksh93-style 
>> interface is to be preferred over any of the other possible interfaces.
> 
> That's an interesting argument to bring to the PSARC meeting.  But clearly 
> that has already happened, so we're a bit late.

I haven't ploughed through the ARC cases, so I'm not sure if this point 
came up or not.

> Clearly, we have not analyzed the run-time behaviour of the new commands; 
> we would have found the bug in ksh93 (~()) and we would have found that
> the ksh93 wants to open ".paths" in every element of $PATH.
> 
> I think we should fix that: we need to do that for the other shell scripts
> also (alias, bg, etc) and it's easier to keep sleep as it now is.
> 
> As for /bin/basename: clearly the optimized version exist for a reason; 
> that is clearly not the case for sleep.  Clearly, making basename a script 
> would be wrong.
>
> But it might be different in other cases, specifically when the built-in 
> command is faster than the standard Solaris command.

Agreed.  What I'm trying to point out is that the justifications I've 
seen to date for this change are not the full picture, and the decision 
to implement /bin/sleep (and others) has to include more factors than it 
apparently has to date.

To be clear: I'm not a ksh93-hater: it is the login shell that I use. 
That doesn't however equate to me believing it should necessarily be 
ubiquitous throughout the OS.

-- 
Alan Burlison
--

Reply via email to