32 bit OS?

Just prior to the Wakefield show, RISCOS Ltd released an announcement stating that they would not be developing RISC OS to become a true 32 bit operating system, despite the fact that one of their leading parts of the mission statement on the website stated that they would be moving to 32 bit in around a year from the initial release of OS 4.

The timing of this statement was terrible, coming slap bang on the Wakefield weekend - an act of either miscalculation I've not seen in years (the last time was the cancellation of Pheobe and prior to that, the leaking of the Pheobe from Acorn - one almost killed the platform, the other almost killed the sales of the RiscPC) or due to not getting the message out in time. It did do one thing though, it seriously dented a large number of both users and companies in the market confidence in the long term viability of the OS.

When I was approached to write this article, I had some deep reservations, but on advice I decided that rather than jumping in with both feet and blaming everything on RISCOS Ltd, I would ask around and see. Given the nature though of this article, I have decided that to further protect the market place, I will not say which companies have provided the statements. In a private email, this has been classed as cowardice, but as I said in the reply, "is it cowardice not to reveal the companies who may end up losing sales due to the comments?"

Firstly, what does not having a 32 bit OS mean? "At the moment, nothing" - if we are to believe what we are told. What it actually means is the stagnation of both software and hardware development. It is not true that the hardware companies have not asked for the 32 bit OS. One of the three main companies producing RISC OS based desktop hardware asked for it two years back and only signed up for OS 4 on this promise. Another hardware company, who, while not directly selling machines, have a vested interest in embedded ARM technologies have stated in the past the need for the 32 bit OS to enable them to use RISC OS with their products - currently, they cannot do this as they would be restricted to using the ARM7500FE or SA110 chipsets. This is not much use when a client requires (say) an ARM9 or X-Scale core.

I would have expected the above from the hardware companies, however, I would not have expected the same back from the software companies, given that any programs purely in assembler or with an assembler component would have problems (especially when the companies have long since left the market - such as Computer Concepts and Sibellius) when moving to the 32 bit OS. The reverse was true.

One of the games production companies still in the market (not RCI) were very forthright in their opinions. The statement back read "The post that ROL made about no 32-bit OS until commitment is shown by the market made me scream. The next day a colleague was going to buy a new computer but due to that irresponsible post he didn't." and continued "If things don't shape up soon then I may have no choice but to abandon all my companies projects and pull out.". Other companies approached voiced the same concern.

In the past, the development of RISC OS has been driven by the need to support new hardware and in particular the need to use new hardware in new machines.

RISC OS 3 was developed because support was needed for the new hardware used in the A5000 and subsequent machines.

RISC OS 3.50 was developed for the new hardware in the RISC PC.

RISC OS 3.60 was developed to provide support for CDROMs and hard drives >512MB and also the original ARM7500 machines.

RISC OS 3.70 was developed to support StrongARM and 3.71 for the ARM7500FE.

RISC OS 4 was originally developed to support the Pheobe hardware.

None of the above were originally intended to be sold as upgrades for machines without the new hardware, although RISC OS 3.10 was sold in large numbers for this purpose.

RISC OS Ltd don't appear to be following this policy. It is understandable that the first release of RISC OS 4 should be sold as an upgrade, since following the cancellation of Pheobe, there was little choice. However, it was always clear that the future of RISC OS is very closely tied to the production of new machines. It is also clear that something more modern than the current RiscPC is needed; that's why Acorn were developing Pheobe after all.

If we are to see RISC OS machines continue in production for many years to come, the question of 32 bit processors will have to be addressed. Many people are leaving RISC OS because they are not convinced that machine development will continue. The lack of commitment to a 32bit RISC OS only reinforces that view.

While there is no denying that the production of the 32 bit OS would be a large task, it has been revealed on a usenet posting from someone at Pace that they did indeed have a 32 bit version of RISC OS which had had normal and typical RISC OS software (such as !pdf and !TechWriter) running on their test bed. This must be one of the most damning statements of all. The OS is available (from what I read), the hardware companies have now gone as far as they can with the archaic hardware we've had to use for ages now (how much more can Castle push out of the Kinetic?) and are now actively looking at 32 bit chipset alternatives.

It would be foolish of me to say that it is a case of the OS is there, get ROS Ltd (or another company) to licence it from Pace for the desktop world and then we can all go and buy the new Evolution or Omega or whatever machines with 32 bit processors, a super modern OS and all the advantages that will give. It negates one thing - money. Pace, like any other company in the world, is, at the end of the day, in it for the money. I cannot see the software engineers there working for months on our favourite OS only to say "here it is guys, do with it as you will and you can buy us a pint at RISC OS 2001 for it". This is where the RISC OS Select scheme should be used. However, it looks to me like nothing more than a few fancy bits missing from RISC OS 4, a pile of bug fixes and a couple of new pieces of kit (such as dhcp).

There is simply no road map from RISCOS Ltd saying where the Select scheme is going. Is it, as it looks like, purely to provide an income to RISCOS Ltd and provide upto 3 sets of unspecified updates or are the end of it, going to see OS 5 as a true 32 bit OS (as Acorn should have made OS 4 for Pheobe)?

In that case, what are the alternatives as I cannot see Castle, Microdigital and RiscStation being able to stump up the large wodge of cash required to pry OS 5 from Pace's mits (and from their antics, I would not want RISCOS Ltd to obtain the licence)? If all three (and any other company with an interest in the 32 bit OS and the future development of our platform) grouped together, it is likely that they would be able to group negotiate with Pace for a decent price, and then through a hardware neutral company (such as Clares - a company with a long track record in the RISC OS market, well respected and honest), sell OS 5. For us with RiscPC and ARM7500FE based machines, this may not be much of an upgrade, but for the new hardware, it would mean that we could move forwards.

A very damning comment (also from usenet) came from the people producing the new Solo machines. They have the choice, a 32 bit chipset which are becoming cheaper as they become more abundant and a free OS such as Linux or the current chipset with OS 4. Given that the end user of the Solo doesn't really care what the OS is, as long as it works, it would not take a genius with a degree from LSE to see which would be chosen. It would be purely speculation at this point to say which would be used, but if we had a 32 bit OS and a discount for the OS 5 licence (something which can be pointed at for being missing at the launch of OS 4, with the subsequent loss of revenue from hard up schools trying to batter off the threat of MS domination), the outcome may be that other developing countries get to see that there are real operating systems out there....

A second factor which should be pointed to is that of hardware abstraction (so called hardware independance). As it stands, RISC OS machines are horribly expensive. This is due to the need for the use of IOMD and VIDC. It's the old rule of "specialised = expensive". Current RISC OS machines need these two chips to work, even the Omega has to emulate these two chips. The alteration of RISC OS to 32 bit would also include this extraction layer with the upshot of hardware prices rocketing down to those of decent (i.e. not from certain companies!) PCs as the OS could be used on any motherboard. While it has to be admitted that the RiscPC, A7000s, RiscStations and Micos do represent a good long term investment, the headline price is what is looked at - who is going to buy a machine costing between £500 and £1000 running at between 56 and 233MHz when for the same amount they can have a relatively good PC with an 800MHz processor and all you would ever ask for (including a very poor OS).

Another factor in pricing is that of RISC OS itself. Look at it this way, a STB OS licence should cost around US$ 1-2 per box (in the order of 10,000 units). For £12, you can pick up copies of WinCE in tens with Pocket Word, Explorer, media player and all the other standard apps. The pricing structure is just too expensive to be used for anything other than a fanatic's OS for an ARM based machine - even Windows Millenium and NT cost less than OS 4 (and you don't need a to buy a browser for those either!). A far more economical alternative would be for Pace to sell the OS prebuilt for a couple of thousand pounds and allow the hardware companies to modify the code for themselves with a royalty going back to Pace. Alright, this way would be probably be the end of RISCOS Ltd, but on the upside, we'd have cheaper machines, have the updates from source and a squeaky clean new OS.

In conclusion, for the most part, I find the statement from RISCOS Ltd regarding the need for the 32 bit OS and their reasons very distasteful, as did a large number of people at Wakefield and who emailed me after my usenet postings on the subject. The non-development of the desktop 32 bit OS really is hammering more nails into the coffin which, until recently, had been removing them with cautious optimism. I make no apologies for my concern for our platform and the operating system we have all come to love and trust and no further apologies to RISCOS Ltd who, since the launch of OS 4 have really done nothing to support or progress the OS.

We need the 32 bit OS and the hardware to match. We don't need schemes to update our aging 26 bit OS when the world goes 32 bit. It's almost as ridiculous as buying an analogue Sky system when they close down that service at the end of June 2001.