The 21-strong ACE Consortium’s new hardware standard is overshadowed by politics
The most significant computer industry announcement of the 1990s took place last week at simultaneous events in New York and Brussels – at least that is how Digital Equipment Corp viewed the launch of the ACE Advanced Computing Environment, which it reckoned to be more comprehensive than Sun Microsystems Inc’s Sparc strategy. Aside from DEC and key members Compaq, MIPS, Microsoft and Santa Cruz Operation, 16 other companies declared themselves on the day. They were the Acer Group, Control Data Corp, Kubota Computer Inc, NEC Corp, Nippon Kokan KK, Olivetti Systems and Networks, Prime Computer Inc, Pyramid Technology Corp, Siemens-Nixdorf Informationssysteme AG, Siemens AG Automation – no doubt included as a separate entity so that the consortium could claim the key to the door – Silicon Graphics, Sony Corp, Sumitomo Electric, Tandem Computers, Wang Laboratories and Zenith Data Systems-Groupe Bull. In fact, of the major supporters of the MIPS RISC architecture at the centre of ACE, only AT&T Computer Systems was not represented. The event set out to mark the establishment of a new specification of a computer architecture based around the MIPS R4000 RISC chip, supposed to act as a base for widespread market support the model being the IBM Personal Computer in the 1980s. But the announcement also included emotive issues such as support for Microsoft Corp’s new technology OS/2 3.0 and Santa Cruz’s Open Desktop Unix-based operating system, issues that dominated attention, and highlighted differences amongst Consortium members that should have remained hidden, at least on the first day of a brand new joint venture.
Advanced RISC Computing hardware specification is the key
Perhaps the most important aspect is not the political operating system choices, but rather the underlying hardware on which the whole thing depends. The Advanced RISC Computing Specification known as ARC – defines a common set of hardware and system firmware interfaces to both operating systems and applications, allowing for the development of a shrink-wrapped base of products that will run on any ARC-compliant machine without modification. It’s basically a fingerpointing problem between hardware and software at the moment, said David Kearney, software marketing product manager at MIPS. But with ARC, if compliant software doesn’t run, then you know the problem lies with the hardware. The ARC specification has been made more flexible by defining hardware interfaces at a higher level above the base register level – and includes a system firmware interface. This allows for greater hardware independence without affecting applications compatibility – necessary when individual Consortium members demand support for a whole range of buses. EISA, AT, Micro Channel, DEC’s TurboChannel and the emerging Futurebus+ standard were mentioned at the conference. It also allows for a more rapid evolution of hardware technologies and a broader choice of compatible systems, the weakest part of the IBM clone situation. The basic elements are: MIPS RISC processor (R4000, or R3000 to start with, with little endian byte ordering); the interfaces to standard input-output devices, such as storage subsystems and network interface controllers; standard 101-key IBM-compatible keyboard; system firmware services, including initial program loading and execution; input-output subsystem functions; and storage media formats. The spec is designed to be scalable, so that it is suitable for laptops or multiprocessors. Defined, developed and evolving under the control of an industry technical committee chaired by MIPS, the specification is available from MIPS to any vendor intending to build ARC specification products willing to sign a non-disclosure agreement drawn up by the plotters.
Intel iAPX-86-based machines are also a part of the specification
Intel Corp was understandably keeping a fairly low profile at the announcements last Tuesday – it can’t be seen to be endorsing the MIPS RISC chip, especial
ly when it has its own 80860 RISC still rattling away in its closet full of skeletons. But nevertheless, industry-standard personal computers could not be ignored, and so they too are part of the deal. To offer compatibility, the ARC spec includes several aspects in common, including the keyboard, storage media and file formats, the consistent little endian byte-ordering and personal computer-compatible input-output interfaces that will make possible the use of thousands of boards and peripherals that are already on the market.
Fragmented ACE does not worry Sun
Sun Microsystems Inc is practically giddy with delight at the way the ACE Initiative came across. According to Sun GSG Advanced Development director Dave Ditzel, ACE could have been a real thorn in Sun’s side if the makings of a single standard and the momentum were there. But with the ACE operating system story as fragmented as it is and even the ACE leadership as lukewarm in its support as Compaq’s appeared to be, Sun reckons ACE will not be a serious competitor. Sun’s performance rivals, he said, will continue to be IBM and Hewlett-Packard, with the DEC RISC VAX entering the arena when it finally makes an appearance. Independent software vendors, he claimed, are going to be hard put to decide which of the three ACE operating systems they should write to. A fourth alternative could also appear, he mused, with Sun’s newly-formed subsidiary SunSoft implementing the System V.4 version of SunOS to the MIPS architecture, if there were any volumes to support such a move. SunSoft of course is already expected to move the System V.4-SunOS to Intel’s hardware. Doug MacGregor, president of Solbourne Computer Corp, the first Sparcsystem builder, was also unimpressed by the lack of solidarity behind ACE. It’s an impressive group of companies all going off in their own directions, he said. What made the Personal Computer work, he noted, was that it was driven by IBM, not a consortium. Consortia haven’t proven that they work. What’s working for Sparc is that it’s driven by Sun. What’s lacking for ACE is that there’s no-one there to drive it… and there are four different environments – with four aces on the table, things begin to smell a little fishy. MacGregor, however, saw a benefit in so well-respected a company as Compaq openly recognising that its MS-DOS strategy isn’t sufficient. This is an opening that Sun and the Sparc cloners will exploit.
Microsoft lifts the veil on what its OS/2 3.0 New Technology is planned to include
The Advanced Computing Environment consortium chose to endorse two operating systems as part of its machine – one not yet implemented on MIPS RISCs – Open Desktop from the Santa Cruz Operation, and one not yet even written – New Technology, or OS/2 3.0 from Microsoft. Both were explicitly said to be supported by all vendors in the initiative, despite the fact that they are basically competing products. Started as long ago as 1988, according to Microsoft, NT-32 or OS/2 3 is to be a high-end 32-bit operating system that can run MS-DOS programs as native, but that also provides a solid base for the forthcoming Windows-32 version of Windows 3.0. The layered software will provide basic operating system functions at the foundation (or kernel) layer, sup-porting protected subsystems (servers) to run as standard user mode applications, providing system services and environments for the applications. There will be subsystems for Windows, MS-DOS, OS/2 and Posix applications. The kernel itself, called the Executive, will be a microkernel of some 50Kb compiled code, aimed at making the system efficient and portable. NT will support symmetric multiprocessing and security to C2 level, including secure distributed processing. It’s due out next year. Open Desktop, from Santa Cruz, should be ready at the beginning of 1992 for MIPS machines, with development versions in the latter part of this year. Extra features will include 64-bit support to take advantage of the R4000 chip; support for the Open Software Foundation’s Distribut ed Computing Enviro
nment and Distributed Management En-ironment software; compliance with XPG4, Posix 1003.2/4/6/7 and Unix System V Interface Definition 3 standards; Motif 1.1, X Window System 11.4 and IXI Ltd’s X.desktop; support for threads, network licence management, logical volume management and diskless systems; and many more, such as secu-rity, hypertext and open systems interconnection support. Initial bus support will be for EISA and TurboChannel. Santa Cruz nowhere mentions System V.4, and appears to lean towards OSF/1. DEC and Bull said they would be underpinnning Open Desktop with OSF/1 at the very least, and an executive with Silicon Graphics said he expected OSF/1 to be the base. However, Jim Wilt of Santa Cruz insisted it was plying the middle ground, and would take both OSF/1 and Unix Labs technology.
Non-disclosure pacts hogtie Apache Unix System V.4 supporters
The Apache Group, those renegade ACE big injuns – big endians that want to run neat Unix System V.4 on MIPS boxes, kept a very low profile last week, as predicted here. And no wonder. Those non-disclosure agreements they signed demand exclusive public support of the little endian principles implicit in ACE as well as exclusive public endorsement of MIPS as their sole RISC system. There’s no middle ground here. It’s either little endian or big endian. You have to choose sides and if you turn coat and go big endian you either have to wait 30 days until after the last receipt of early vers-ions of the ACE specification to say so publically or give five days notice that you’re changing sides – whichever is longer, apparently. Oh yes, and you have to return your copy of the ACE specification…
V.4 Applications Binary Interface crew
AT&T Co’s Unix System Laboratories Inc couldn’t really have picked any other day to announce the availability of its System V.4 Applications Binary Interface for the MIPS RISC chip, supported by NEC, Olivetti, Prime, Pyramid, Siemens, Sony and Tandem, as well as AT&T Computer Systems, pretty much the only significant MIPS vendor not declared as an ACE consortium member. The applications interface of course uses the standard Unix big-endian form. Unix Labs also announced its System V.4 reference machine, based on Pyramid’s MIPS System V.4 operating system, with significant contributions from NEC and Sony Microsystems, would be available on a Sony News 3200 Series laptop by the third quarter.
Other operating systems will appear for the hardware, predicts MIPS
Although MIPS Computer’s David Kearney predicted that most ACE systems would go out with either NT or Santa Cruz Open Desktop as the operating system, he said that other operating systems were more than likely to appear on the ARC hardware in the near future, as companies see a market opportunity in following the spec. This might be one route that Unix System V.4 supporters take, enabling them to provide V.4 systems without the Consortium as a whole having to endorse it. It could also be an attractive route for smaller companies with proprietary operating systems such as Atari Corp and Commodore International Ltd, and, said Kearney, for niche operating systems such as the Lynx/OS real-time system, which already runs on MIPS hardware.
Widespread scepticism remains over ACE
Research house International Data Corp gave the ACE announcements a mixed reception. While saying that the initiative was the first building block in the foundation of a potential new standard for networked, distributed desktop computing, the organisation went on to say that many cracks could develop, and it is premature to call the ACE initiative a success. Although the consortium could slow Sun’s momentum by grabbing the attention of software developers, users and system vendors, says IDC, Sun still faces a more serious threat from IBM and Hew-ett-Packard. Individually, IDC’s Vicki Brown said it will be difficult for a consortium of companies to compete against one company, namely Sun, that has its act together, while David Card commented that the unanswered question is what is common be-ween Unix
and OS/2. If you’re a software developer, you still have to choose one or the other.Will ACE duplicate the success of the IBM Personal Computer? The main difference is that the PC standard grew from market demand rather than pre-ordained marketing, and few observers expected the scale of growth. In 1981, Dataquest predicted that the market for desktop micros would rise from $1,300m to $4,000m by 1985, and that IBM would win some $700 to $800m of it. In the event, the market turned out to be five times as large, with IBM winning $5,000m by 1985. Even Sun Microsystems, which has also been credited with inspiring the ACE consortium, only gradually realised it had a chance of creating a similar workstation standard through its Sparc strategy, as its previous Motorola and Intel product lines testify. The ACErs are anticipating such a demand before it has shown any sign of existing.
John Abbott & Maureen O’Gara
