One question that nearly every UniForum member has an opinion on is UNIX unification. It is an issue whose flames have been fanned by the inception of each systems vendor's brand of UNIX, and it was ignited again last March with the reorganization of the Open Software Foundation (OSF).
When we asked members the question: "What is the next logical step for UNIX unification?" you came back with answers even more diverse than the UNIX flavors themselves. The answers did not reveal a consensus on which direction UNIX unification should take, but rather three distinct ways of looking at the issue. Many look for the solution in the realm of multivendor standards bodies and their processes and brands. Others put the burden for bringing the industry together on the vendors themselves. Still others see the answer in purely technological terms.
Don Mason, of Eagan, MN, came out clearly in favor of the standards bodies: "Continued clarification of the roles of various collaborative organizations such as OSF, X/Open, and UniForum is needed," Mason said. "Common specifications such as Spec 1170 and the Common Desktop Environment (CDE) are good starting points for addressing UNIX weaknesses. They will need ongoing work; a one-shot Spec 1170 doesn't seem viable in a changing computing environment."
Claudio Vaccarella, of Rome, Italy, says "Concentrate the standardization process under X/Open." And not surprisingly, X/Open's director of marketing communications, Jeff Hansen, agrees. The next step is "formal branding of products by X/Open, a wide variety of products carrying the X/Open brand, and the incorporation of branded UNIX by users in their procurements," Hansen says (see related article in this issue).
David A. Trevino Rodriguez, of Monterrey, Mexico, also advocates further standardization. The industry needs "unification on COSE and X/Open," says Trevino Rodriguez, "and in the future, an open road to objects."
But the unification road has already taken a scary turn for Glenn Kapetansky, of Naperville, IL. "It looked like X/Open/ COSE/OSF had finally combined on the standards end and Novell was holding the torch," he says. "Now that Novell has sold a perpetual license to Sun, I worry about a re-divergence of vendor products."
Standards are the answer for Kim L. Shiveley of Richardson, TX, who advocates adhering to POSIX. "Spec 1170 is a step backwards for those who were adopters of the POSIX standards," Shiveley says. "Use COSE as an umbrella to approve portability standards. COSE and CDE should not be writing new supposed standards." Shiveley gets agreement from Mike MacFaden, of Fremont, CA, who says POSIX and Interprocess Communication and Synchronization (IPCS) should be implemented "across all major flavors of UNIX."
One who would get the standards bodies out of the way is Jerome B. Senturia of Cleveland, OH, who says, "OSF was originated to slow the AT&T process of UNIX unification. It succeeded so well that Novell now owns UNIX (or is it X/Open that owns it?). OSF should be laid to rest and System V release 4 and its relatives should be left to the commercial marketplace."
Equally frustrated is L.A. Lundquist, of Austin, TX. "COSE was to be a giant quantum leap, yet it is so bogged down that it can't get even the first entity out in rapid fashion," Lundquist says.
Clearly putting the burden on vendors is John Modransky, of Hoffman Estates, IL. "Have vendors using UNIX as a core operating system drop the pseudo-names such as HP-UX," Modransky says, and beyond that, produce "true binary compatibility across like CPU architectures and true source code compatibility. The only way to lure ISVs to UNIX is to make it easy to create applications that do not have to be altered just to run in another UNIX environment."
Rudy White, of Moffett Field, CA, is pessimistic that vendors will ever do the right thing. "I would like to see a non-proprietary UNIX-a truly open UNIX-but that is not going to happen," he says. "If the big guys fall or even become threatened, they will join forces and become a powerful consortium. Most users aren't sophisticated enough to see the real beauty of open systems and the manufacturers aren't stupid enough to let it happen."
A.F. Gualtieri, of Long Grove, IL, has this piece of advice: "Vendors should deliver products that follow standards and stop insisting on their own version," he says. "The users will buy."
Still suspicious of Novell is Barry E. Hedquist, of Santa Clara, CA, who says, "There is only one real way to achieve UNIX unification, and that is for someone to achieve and maintain market dominance while providing leadership to the UNIX community. USL was on its way to doing so with System V release 4 until being sold to Novell. Now it is anyone's guess as to what will happen. Novell has sold the UNIX name to X/Open but is keeping the source code. Is this logical?"
Some members see the unification issue in technical terms. Jay A. Krone, of Marlboro, MA, zeroes in on the issue of the sockets interface employed by Berkeley-derived UNIX operating systems, and its incompatibility with the streams interface used by SVR4. Programmers attempting to write C code that will run on UNIX systems must write for one or the other interface. "As long as the Berkeley-derived UNIXes with market presence-HP, IBM and Sun-use sockets and most everyone else licenses SVR4 with streams, HP/IBM/Sun will have a hold on the market like IBM did in the old SNA-vs.-everything-else days," Krone says.
Compliance with standards should be tested with a set of verification tools that will allow providers of software, hardware and peripherals to become certified, says Brad Bright, of Irvine, CA. "Without some consistent way of knowing if you comply with the standards, it is hard to have a real plug-and-play environment," Bright says.
"Drop baggage from the past," advises Liz Farrell, of San Jose, CA. Remove backward compatibility and allow fewer commands. Lose command line capability and force users into a window-only interface. Limit the ability of the user to use freeware."
For Bruce McIntyre, of Mount Laurel, NJ, "The next logical step is a complete system call common interface, and then a portable binary interface. I really don't care which one, as long as it's one."
Bill Rieken, of Los Gatos, CA, suggests, "It would be nice if UNIXWare and Solaris would use the same device interrupt settings. When a customer buys a notebook which can load Windows and SCO UNIX from CD-ROM, but it cannot load UNIXWare or Solaris, the words open systems are exposed for the joke they are."
And finally, David McCall, of Rohnert Park, CA, suggests, "AT&T and Berkeley must elope and produce offspring which is better than the parents. Or perhaps they both must just be replaced by something not yet known."
The next step, logical or not, may well coincide with one or more of these suggestions. And its importance for the future of open systems will be not which path is taken, but whether it leads toward unification or away from it.