Multi-instruction set chips are key to portability

Tom Mace, president and chief executive officer of 88open, believes the proposed Architecture Neutral Distribution Format-type software portability technologies now doing the rounds should be pursued as vigorously as possible, but argues that provision for support and verification must be developed, if the stuff is to become accepted by a still sceptical independent software vendor community. Because the techniques involved imply a separate implementation of the portability software on each different system, they also mean that what users purchase and what they see on their screens are two different things, there are obvious questions of support and testing that need to addressed. Further out, Mace believes these portability issues will become redundant as chip technology evolves to the point where multiple architectures can be implemented and supported on a single microprocessor. Although there are some formidable technological hurdles to overcome, the concept is already being put into practice in several projects, most prominently from where Mace stands, in the IBM Corp-Apple Computer Inc-Motorola Inc effort to bring some of the functionality found in Motorola’s 68000 and 88000 processor families onto the PowerPC RISC architecture that the three are developing, as well as maintaining binary compatibility with IBM’s RS/6000 Power RISC. There will be fewer people and fewer distribution issues involved in these efforts than in the Architecture-Neutral Distribution Format-like developments, argues Mace, who reckons real on-chip multi-environment implementations are some four to five years away from commercial realisation. Although IBM, Apple and Motorola have a forged an agreement for the duration of their project, there remain obvious legal implications for a vendor that might want to stick say Sun Microsystems Inc Sparc and MIPS Computer Systems Inc R-series instruction sets on to a single piece of silicon. Mace believes those copyright and interface issues are now being worked through in the current round of legal skirmishes between Intel Corp and its iAPX-86 impersonators such as Advanced Micro Devices Inc. The upshot, Mace hopes, is that interface copyrights will be broken and will therefore pass into the public domain, and that subsequently, whoever can produce the cheapest, fastest part with the most capabilities to mimic other architectures will win out, and once one does it, the others will be forced to follow.

First iteration of PowerPC to be 30% less powerful than the 88000

On the PowerPC effort, Mace says the initial part will come in with around 30% less power than the 88000, though he expects performance to converge over time. PowerPC, which Mace describes as not elegant, but a good start, incorporates around 15% of the 88110’s architecure, re-engineered of course. Meanwhile the 88110 itself is still available only in developer and very limited edition quantities, despite announcements from the likes of Dolphin Server Technology A/S and Harris Computer Systems. Motorola, Mace says, is still cleaning up the part and doesn’t want to announce pricing or general availability on it until later in the year.

88open test suite for System V.4 on the 88000 RISC

The 88open Consortium Ltd has released a compatibilty test suite for applications – in binary and object forms – running under Unix System V.4 on the Motorola Inc 88000 RISC. Software developers – independent or corporate – can have access to the tests and certify their System V.4 applications on the 88000 at no charge, the group says. The Posix and X/Open Portability Guide 3-compliant ACT/88 for System V.4 adds tests for new Unix V.4 features such as dynamic linking, Extended Language Facility and streams.

New adherents to 88000 RISC seen as system sales exceed $1,000m

Tom Mace claims that some $1,000m of business has now been conducted on 88000-based system products, and claims, as the group often does, that there will be more converts to the cause – this time around there are five or six manufac

turers using the Intel Corp 80860 RISC on its hit list. 88open will open a European office this summer, either in the UK or Germany, which will use space rented in one of its members’ offices. The office will co-ordinate 88open’s pan-European marketing, information, exhibition and conference efforts. The group’s clutch of porting and testing centres remain intact.

System V.4.2 desktop Unix too expensive to win

Tom Mace was able to cast some interesting light on the heritage of Unix System Laboratories Inc’s Destiny desktop Unix – now going out as Unix System V.4.2. He reckons that up to 70% of Destiny originally featured in the modularised Unix System V.3 implementation that was developed by Convergent Technologies Inc – the company now subsumed into Unisys Corp – for the ill-fated Unix personal computer that it manufactured under an OEM agreement for AT&T Co, one of the many failed initiatives tried by the telephone giant before it threw in the towel and decided to buy in commercial computer systems management wholesale by acquiring NCR Corp. The stand-alone desktop machine which appeared way back in October 1984 (CI No 42), was a Motorola Inc 68000-based machine that came with 512Kb RAM – expandable to 4Mb – 20Mb (Winchester) drive, 5.25 floppy drive and 12 mono display with a telephone and modem built in. The thing retailed at between $5,000 and $6,000. Although at that time the Unix personal computer was the most successful Unix product ever, by May 1986, AT&T was struggling to market thousands of the things held in stock – it had wanted to sell a million of them – and had co-opted Convergent on to the job too. The problem then, as now, argues Mace, is that Unix systems are developed by Unix engineers who like the complexity of Unix. If you develop something complicated, then you need complicated people. The problem with Destiny is that even at $350 – which will only buy kernel-level stuff – it is still much more expensive than MS-DOS with Windows, and Microsoft Corp is sure to pitch Windows New Technology at some attractive price point well below that of Unix Labs. Mace believes that while Unix Labs may be able to claim the value added in Unix is worth it, the problem is who wins by saying things like that? Only Apple perhaps. – John Abbott