"Unix is the operating system written to run a space game; every vendor has a different version of Unix; you can't transfer a Unix program from one Unix system to another without major rewriting; Unix is twenty years old and was never intended to be used outside of a laboratory; Unix isn't suitable for transaction processing; the security holes in Unix make it impossible to use in a commercial environment". I have heard every one of these statements from mainframe users in the past couple of years, and they are all solidly based in fact. They are also false, and IBM has just admitted this.
A lot of the arguments against Unix have been heard from salesmen in every division of IBM. The AS/400 in particular has competed against the swathe of mini manufacturers, all offering Unix systems of one kind or another, for a long time. The mavericks of the PC division long ago adopted Unix as a secondary line, with first PC/IX and then AIX, to the embarrassment of the AS/400 men. Only the mainframes have, until now, been impervious to the challenge of Unix.
The counterarguments for Unix say that you can set up an expandable, distributed strategy for IT based on Unix systems for much less than any non-Unix (they say "proprietary") solution. Into your system you can connect hardware from almost any supplier, hire Unix experts from almost anywhere, buy any of thousands of applications to run on your system, and generally do it faster and cheaper. Again, this is partly true and partly false.
Much of this success has been due to the informal adoption of standards by users and manufacturers of Unix systems alike. POSIX, X Window, Motif, the C language, Ethernet and TCP/IP are all broad based standards that have evolved from work with Unix.
IBM have always put forward their own internal systems as standards to be adopted by lesser manufacturers. This was true with SNA in the '70s, and with SAA and Token Ring in the '80s. Sometimes, as in the case of the original IBM PC architecture, this has succeeded. Most of the time it has not. There was a time when it looked as if SAA would become a real standard through its use in Windows and OS/2, but that chance has now passed, due, in no small way, to IBM's own inability to conform to its standard.
For multi-user systems, Unix has been dominant in the small to middle end of the business since the mid '80s, although it crept up on us quietly, and we have been slow to feel its presence. Only mainframe users have kept apart. The one lone exception in the market was Amdahl, who quietly sold in increasing (but small) numbers its own Unix implementation, UTS, to American government and educational users.
IBM traditionally issue a mountain of press releases each September. SAA was first announced in one of these pagodas of paper, as was AD/Cycle. Last year SystemView was the star of the 1991 model year. This year, IBM has quietly announced that their in-house version of Unix from the PS/2 and RS/6000 product lines, AIX, will run native on ES/9000 and ESA/390 mainframes.
AIX is based on the Open Software Foundation's OSF/1, as is also the Open Desktop implementation to be developed by the ACE consortium. OSF/1 supports the POSIX standard from the IEEE, a standard for an application programming interface. Programs written according to the POSIX standard will run on most Unix compatible operating systems, as well as a number of others. IBM has stated that they will support POSIX in their MVS and OS/400 operating systems. The industry standard X Window System will also be supported in AIX/ESA.
IBM have said that AIX/ESA "provides an IBM solution where 'mainframe UNIX' is a requirement". This is a reference to the US Government's requirement for POSIX support in many of their tenders. However, related announcements will also mean that the mainframe can be used as a file and compute server for work stations. TCP/IP, NFS and Ethernet options will be supported in this environment, allowing the vast realm of work stations (from Sun, NeXT, HP, DEC and others) that support these protocols to access IBM's big boxes. TCP/IP will also be supported for OS/2 and PC-DOS systems.
This indicates a more "Open" direction from Big Blue, following earlier announcements that Systems Application Architecture (SAA) will support certain AIX facilities.
Why should IBM make such a complete about face? I think that the weak state of the global economy has a lot to answer for. IBM's financial results have been very poor of late, and they are making alliances as if they don't know which way to turn. The changing fortunes of the alliance with Microsoft, the new operating system plans with Apple, and the "Business Partners" involved in the SystemView strategy all show a certain desperation to ally with anybody who offers.
SystemView has passed out their core system software to third parties, and the "Pink" project won't have a production release for at least three years (at which point it will have no more than NeXT already have today).
So it is no surprise to see IBM suddenly accept all of the Open Systems standards that she has been resisting for so long. However, I believe that the situation is more desperate than that. In the UK there has been a noticeable reduction in the numbers of mainframe sites, as companies consolidate and down-size operations. A standard desktop PC is more powerful than a typical mainframe of ten years ago; organisations who take care of their budgets realise that they can replace, for less money, a mainframe with a network of minis and PCs. For this solution, the Open Systems approach is the only viable route.
This decision has a lot of implications on the rest of IBM's strategy. The AS/400 is shown as a very poor contender. Connectivity for the AS/400 is almost non-existent, applications are not portable (and so are more expensive), and price performance has only been maintained by discounts that have closed down many AS/400 resellers.
The RS/6000, with its AIX version of Unix, now looks like IBM's strongest offering. It can validly fit in to a companies Open Systems strategy, and looks quite competitive in price, although only only over a long timescale.
The Open Systems package from IBM is just a tentative move by that largest of large organisations away from their mainframe strategy; but once they have their foot on that icy slope, there is no stepping back.