[illumos-Developer] webrev: printf

[Garrett D'Amore]

On Thu, 2010-10-14 at 16:32 +0200, I. Szczesniak wrote:
> On Thu, Oct 14, 2010 at 6:08 AM, Garrett D'Amore <garrett at nexenta.com> wrote:
> > So I finally got tired of waiting for someone else to do a printf
> > replacement-
> 
> So why is this a surprise for you that you have difficulties to find
> contributors? I'm not really surprised. I've personally invested time
> and resources to give Illumos a standard conforming tail
> implementation but was turned down even after issuing a commitment to
> the job and already finished code and standards testing.

I've explained repeatedly that we needed a /usr/bin/tail asap, and that
the code I had from ksh93 was not up to the task.  Chris Love stepped
forward to do the work *before* anyone from the ksh93 team, and it works
well.  HOWEVER, both he and I have indicated that we're willing to
replace his implementation by a ksh93 the first time it makes good sense
to do so.  So far I've not heard a compelling argument in favor of that.

I'll even make the same assertion for printf.  If a ksh93 derived
version makes sense, I'll take it.  I even *tried* to take it.  But I'm
not going to break a zillion scripts because of some philosophical but
otherwise useless argument that printf needs to have option parsing,
especially when the open group lacks any statement about such option
parsing.

In case you've not realized yet, I do not subscribe to the school of
thought that the only reasonable implementation for POSIX utilities is
ksh93.  If you expect illumos to automatically defer to ksh93 for
everything POSIX related, you're going to be disappointed.  That said,
the very instant that the ksh93 folks demonstrate to me that they have a
superior product to the alternatives, I'll happily support their
integration into illumos.

> 
> After this incident it is very hard to convince my own management to
> invest anything in Illumos, at least until Illumos can make clear
> management decisions, clear duty assignments, i.e. who owns which
> code, and a clear commitment to the POSIX standard. All of this was
> not done for tail and the last one certainly not for your printf code.

As I've yet to see any actual useful contributions towards illumos from
the ksh93 folks.  So, after waiting months for clear contribution, this
statement from you hardly seems like much of a loss -- I'll take working
implementation from folks who are working and committed to providing
what we need now, including supporting illumos requirements, over
much-vaunted ksh93-derived vaporware.

I've been asking for a printf replacement (as one of the 'easy' tasks
the ksh93 folks could do) for a couple of months now from the ksh93
folks.  My guess is they could have done this in about 20 minutes, maybe
an hour tops.  It only took me two hours to add the %n$ parsing to
FreeBSD printf and integrate it.

Bluntly, this community is a meritocracy.  If you want to have much to
say here, then you need to demonstrate merit.  So far all I've seen is a
lot of talk and no delivery.

	- Garrett