On Jul 25, 2005, at 1:12 PM, Keith M Wesolowski wrote:
On Mon, Jul 25, 2005 at 03:48:52AM -0700, Roy T. Fielding wrote:
Why does it have to be 100% compatible? That is a serious question.
What breaks so bad that not having access to the source is considered
a viable solution?
100% compatibility is not always required. Sometimes, no
compatibility is required at all. See the interface stability
taxonomy in chapter 7 of the developer's reference. ksh,
unfortunately, is Stable.
ksh93 is pretty darn stable, if you ask me. 12 years stable.
Are the people who write scripts for the sake of portability
refusing to support the Solaris platform because it is stuck
with an ancient, non-standards-compliant copy of ksh? Or do
they just ignore Sun and move their business elsewhere?
There comes a point in any business wherein compatibility with
the future becomes more important than compatibility with the
past, and I'd say that point was passed for ksh about 9 years ago.
The fact that Solaris failed to keep up with the times should not
be considered a good thing -- Sun loses customers because it is
tremendously difficult to administer a system wherein 1 out of 3
system utilities have to be replaced by their more modern open
source equivalents before you can start building applications.
Being worried about scripts written by people running older
versions of Solaris is a Sun business concern that does not apply
to people who have yet to install SchilliX. It is, in fact, of
far greater value for SchilliX to support all the users who
want to migrate from Linux and have a bunch of ksh scripts
written for ksh93. Though, quite frankly, this whole discussion
is a bit obscure because I don't know anyone who writes ksh
specific scripts (just sh scripts that happen to work in ksh).
Of course, that too isn't a straightforward problem because
some Linux distros use pdksh and other use ksh93. See, e.g,
http://www.dbforums.com/t1163962.html
but wouldn't it be nice if we gathered together all of the
remaining OS groups and managed to all agree on ksh93 being
the only standard from that point on?
Nothing I said should be interpreted to mean that not having source is
acceptable. You're overlooking the point of this thread: that neither
staying closed nor breaking compatibility are viable options. The
right solution is to do the necessary work to reimplement sufficiently
compatible functionality that can be integrated into OpenSolaris -
which means open source.
No, the point of this thread is that "viable options" are
dependent on your existing customer base, which is a business
decision. Compatibility is a great thing, but it is not the
only concern.
If there is a serious compatibility issue, then Solaris can replace
the new executables with ones that are 100% backwards-compatible.
There is no reason for OpenSolaris to be so hobbled.
Compatibility is a value that Solaris engineers, users, and
third-party developers share. This community and the Solaris
community overlap almost completely; therefore it is unlikely that you
will find support here for destroying compatibility. That the
contents of a putback can be made open is a necessary *but not
sufficient* condition for their suitability for inclusion.
Inclusion in Solaris is not the same as inclusion in OpenSolaris.
Let's be frank: Sun will be investing a good deal of effort in
replacing some of the closed code with open equivalents. This process
will not be completed overnight, as much as we might like it to be.
Nor does it have to be entirely done by Sun.
In a perfect world, it would have been done before launch; as it is,
we worked our asses off to make available what's there today. But not
only Sun should participate in this process; elsewhere in this thread
someone has helpfully posted a list of ksh93 incompatibilities. The
ksh93 sources are available to the world, and a motivated individual
with access to both those sources and that list of incompatibilities
could certainly create a ksh93 that, when invoked as ksh, operates in
the same way as Sun's ksh88. Following review, that ksh would be
added to OpenSolaris and would appear in subsequent releases of
SchilliX, Solaris, and any other distributions. All this development
could be done in the open, like any other project.
That is what Jörg was asking for at the beginning of the thread.
It was the statement that we couldn't possibly *start* such an
activity until after it had been approved by Sun's ARC process
that caused me to enter the discussion.
....Roy
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org
