On Dec 8, 8:51 pm, Tim Smith <reply_in_gr...@xxxxxxxxxxxxxxxx> wrote:
> On 2007-12-08, Roy Schestowitz <newsgro...@xxxxxxxxxxxxxxx> wrote:
>
> >http://advice.cio.com/esther_schindler/should_ibms_os_2_be_open_sourced
>
> This quote from the article perfectly sums up the OS/2 experience:
>
> At a Comdex in the late 90s, I was in the press room chatting with three
> other computer industry journalists who had also written quite a bit
> about the OS. Jerry Pournelle wandered by. "Hey, Jerry," said Wayne. "We
> were just talking about OS/2." Jerry nodded and said, "Yeah, OS/2. Great
> operating system, with only one problem--"
>
> And we all shouted in unison, "IBM!"
The big problem with OS/2 was that it was developed in the days when
John Akers was saying "If we Build it, they will come", or more
accurately, "If we Build it, they will buy". It was the worst kind of
hubris. In late 1991, MVS 4.0 was released and the anticipated orders
did not come. Over most of 1992, IBM had to lay off a substantial
portion of their workforce. Many predicted the demize of the
Mainframe.
OS/2 2.0 suffered a similar fate when it was released for General
Availability in March of 1992. It was riddled with bugs inherited
from Microsoft, and enabling certain features that improved
performance had a horrible toll on reliability. In 1994, OS/2 2.0 was
replaced with Warp 3.0, which had a "spare tire" mode, allowing users
to generate a recovery desktop. This allowed them to recover from a
crash or disk failure very quickly. Still, it was a messy recovery.
When Warp 4.0 came out, it was very solid, very reliable, and on the
rare instance that you did get that pesky race condition indicated by
the TRAP 0D or TRAP 0E error, it was possible to reboot into the most
recently saved desktop. In most cases, a competent administrator
would save the desktop so that OS/2 could recover in about 10
minutes. Warp 4.0 could go over a month without crashing, far
superior to Windows 3.1, Windows NT 3.1, or even Windows 95.
Microsoft wouldn't achieve comparable reliability until Windows 2000
with SP2.
> OS/2 Warp (especially version 4) was a great OS--in 1994. But does it
> have anything all that interesting in 2007? Is there anything in it
> that would justify the effort it would take IBM to make it open source?
IBM has already made much of the functionality an Open Source
contribution to Linux,
this includes the queue based scheduler, which it also uses in it's
Mainframe and AIX operating systems. They have also included TCP/IP
interfaces to 5250 and 3270 terminal emulation, which works quite
nicely with most modern mainframes.
Many of the other popular features were incorporated into Eclipse, and
Java.
Most of the other functionality was already provided by Linux. Named
pipes, named message queues, and sophisticated networking capabilities
were also provided by improved TCP/IP support on the mainframe, as
well as supprot for MQSeries and WebSphere Message Queue. Linux also
supported MPI, CORBA, and Message Broker, which made it much easier to
migrate from messages generated by COBOL and CICS or IMS to MQSeries.
Even the EBCDIC/ASCII could be pretty reliably mapped using MQSeries
and Message Broker or WebSphere Java wizards.
> And that would be a fair amount of effort. Even if it turns out that
> IBM does own everything in it, they would have to make sure of that,
> which would require auditing a lot of code.
More importantly, IBM just went to war against SCO for bogus claims
that IBM had perloined AT&T System V code and gave it to Linux.
Nearly all of the claims were shown to be groundless, and those claims
that IBM did not ask the judge to dismiss, IBM can prove that IBM
OWNED before they were integrated into UNIX or Linux.
IBM probably doesn't want to get into another FUD war with Microsoft
over OS/2. Even though Microsoft settled by giving IBM OS/2, there
were lots of nondisclosure agreements related to the orginal OS/2
code. Without a blanket agreement from Microsoft granting IBM all
rights, it's unlikely that IBM would consider OS/2 a good candidate
for source code release.
As suggested before, IBM has released Open Source technology which
emulates most of the desired OS/2 functionality. Eclipse, Jikes,
Java, X3270, TN3270, X5250, TN5250, and other "Host Access Technology
Services" (HATS), are available in Open Source or for nominal fees.
Even the MQSeries Client is available free of charge to Linux users,
and allows Linux clients to connect directly to Mainframe MQSeries
servers.
IBM even supports an LDAP interface to RACF, which allows Linux users
to integrate directly with IBM security using Pluggable Authentication
Modules (PAM) using the LDAP option. Access to the LDAP server can
further be secured by using the Kerberos option. If Windows Active
Directory is patched into "Standards Compatibility" mode, Windows
clients can also be integrated, but most Mainframe administrators
prefer to have either Linux or AIX functioning as a protection layer
between Windows as the "real" mainframe applications.
Portable languages used in OS/2 such as REXX have also been ported to
Linux. The Linux version even has features that were not available or
practical on OS/2 or Windows.
> In general, it is not easy
> to open source a large project that was conceived from the beginning to
> be closed, and developed for a long time under that assumption.
That is the big problem. Most of the key technologies have been
published in Open Source form. Not necessarily the original code, but
more often, functionally similar implementations done in Java,
Portable C++, or portable languages like REXX.
In short, there is very little need to release OS/2 as Open Source
Software. Most of the core functionality has been offered in either
the Linux Kernel, the Linux Libraries, or in Applications such as
Eclipse, MQSeries, Java, Jikes, or WebSphere. There is very little,
that hasn't been done better by Linux.
|
|
