Path: keelhaul.demon.co.uk!not-for-mail From: Jonathan Marten Newsgroups: comp.sys.acorn.misc,comp.sys.acorn.advocacy Subject: A radical new direction for Acorn [long] Followup-To: comp.sys.acorn.advocacy Date: 5 Jul 1998 08:29:19 +0100 Organization: The Self Preservation Society Message-ID: NNTP-Posting-Host: localhost.demon.co.uk X-Newsreader: Gnus v5.5/XEmacs 20.4 - "Emerald" These are my (rather verbose, unfortunately) thoughts on the current Acorn products and where they should be heading in the future. They may sound quite radical, possibly even heretical: however, although I have been an Acorn enthusiast ever since the early BBC days, and admire their products and philosophy, I fear that the current product line could be their last desktop computer. Read on to find out why, and what they could do instead of slowly but surely heading into oblivion. The problem right now ===================== ...is that current Acorn desktop hardware is proprietary: whatever plans Intel may have after acquiring StrongARM, it will never become a mainstream desktop or server product. ARM Ltd. admit as much: their processor and peripheral chips are aimed squarely at the low-power portable appliance market where they could reign supreme. It is also expensive (at least in terms of up-front purchase cost): a reasonable Acorn system is two to three times the price of a PC from a box shifter, and comparable in price to a high-end Intel- or Alpha-based system. This is entirely due to economies of scale and mass production, which a small company with a limited market simply cannot match. Advocates may argue pounds and MIPS figures all day long, but it has to be faced that the advantage in raw computing power that was such a revolution when the Archimedes first emerged (ten years ago) is no longer the case, and probably never will be again. However, Acorn's major strength is RiscOS, specifically its friendly and responsive desktop environment. When this first became available there was absolutely nothing else that could match it: however, rival operating environments are now available incorporating many of its good features but with a far more impressive tick list to appeal to purchasers. However, RiscOS still manages to retain the responsiveness and power of those early days, in contrast to the Windows family which requires annual increments in computing power and memory simply to stay still and to run the latest applications. It is still a pleasure to use the desktop system and its applications; they are also rugged and tolerant of users' mistakes (a great advantage for schools and unskilled users), are resistant to viruses and require very little maintenance. In terms of technical innovation, other systems - and I include Windows and the X Window System among those - could learn a lot from RiscOS's simple but powerful font, sprite and printing facilities, as well as the well-integrated Filer and its simple drag-and-drop data transfer. Despite these advantages, RiscOS is showing its age. Leaving technical problems aside, it will remain a curiosity unless it is able to run leading edge applications - this doesn't just mean the latest version of MS Office, but also others such as Netscape, Autoroute, Mathematica, Xara and Tomb Raider. RiscOS may possibly be able to run these, or something very similar, at some time in the future: however, by then the rest of the world will have moved on and there will be another generation of applications and games for Acorn users to wish for. Some dedicated people have ported major open-source applications such as Perl, Ghostscript and GCC to RiscOS, while others have produced freely distributable applications of their own such as Zap and Acornet. I admire the dedication and the ability of these people, and the Acorn world would be much worse off without them. However, there are some archaic limitations in the operating system which make software development an order of magnitude more difficult than it should be, ensuring that even the latest versions of these applications are still several steps behind those that the rest of the world can run. RiscOS still has a lot of historical limitations that show its simple origins but now simply get in the way: 77 files in a directory, short file names and no extensions, limited length of command line without terrible and unreliable hacks, and limited support for multiple processes and shared libraries. It also has very limited facilities for complex shell scripting and unattended maintenance, and almost no provision for remote access, multiple users or security. The filing system (especially the pathname syntax) and the command language are both idiosyncratic and lacking in modern features. The Acorn platform is only very slowly acquiring technologies and applications that the rest of the computing world now takes for granted: Java/JavaScript and Web browsers which support them are only just now beginning to appear, while Real Audio/Video and Net Phone will probably never do so. There is no operating system support for either SMP, PMT or VM: advocates may again endlessly debate the necessity or practicality of these, but for some tasks and applications they are absolutely essential. The hardware, although powerful at its core, is very limited in expansion capability: to even get more than one serial port requires an expansion card to be fitted (even then having limited software and application support), while there is no USB or accelerated graphics even in Acorn's just-announced new hardware product. [Aside: When it was first announced, it was claimed that the RiscPC was infinitely expandable and proof against obsolescence. Much the same claims were made for the BBC (with its "Tube" processor expansion bus) when that machine was announced, but in both those cases it is now evident that that was an impossible promise to keep.] Why Acorn should abandon their proprietary core =============================================== ...because they have to get themselves out of their perceived niche: proprietary and expensive desktop systems not scalable to server tasks, lagging behind in both hardware and software technology but enthused over by anoraks and schoolchildren. At the moment Acorn and the developers in their market have to do absolutely everything: design, build, manufacture and market the hardware, develop and maintain the operating system and the standard applications, provide tools for both in-house and third-party development, and integrate new technology as it becomes available. Applications, development tools and Internet access all depend on facilities provided by the operating system, yet while they wait for much-needed upgrades to RiscOS they will continue to lag behind. Observe the standard desktop applications (Edit, Paint and Draw) for an illustration of this: although these are the applications that a new user of RiscOS will first see and use, they have hardly moved on since they first appeared nearly ten years ago, and it has been left to dedicated enthusiasts to produce improved versions. This is not a sign of inability or intransigence at Acorn, but simply an indication that the massive effort necessary to maintain and develop the hardware and the RiscOS core leaves no time for its peripheral areas. We Acorn enthusiasts are convinced of the power and versatility of RiscOS. But the operating system and its desktop environment only runs on Acorn hardware: anyone wanting to try it out has to first make a substantial investment (a few hundred pounds for even the cheapest second-hand Acorn system), so the decision to do so will not be taken on a whim. Whereas interested hackers and those graduating from Windows can currently try out Linux or BeOS by simply downloading and installing it, or in the worst case by spending a few pounds on a CD: at that price many more will be inclined to try out the alternatives and hopefully stick with them. Insisting on sticking to proprietary standards, against all the acknowledged market forces, completely killed the IBM mainframe business. Digital and ICL almost went the same way, but managed to recover by changing their emphasis to open systems and industry standards. Acorn has to do the same. A new direction for their future product line ============================================= [Note: for clarity from here on, "RiscOS" refers to Acorn's current operating system running on ARM hardware. The alternative "R***OS" refers to what I believe ought to replace it.] Acorn should concentrate all of its development and marketing effort to take advantage of its major strengths: the friendly and powerful RiscOS desktop, its specialised knowledge and established position in the UK educational market, the "enthusiast" image of its followers and the growing aversion to Micro$oft/Intel. Which is why they need to abandon the policy of developing and building proprietary hardware to deliver that. Doing so should be a means to the overall strategy of selling software and the R***OS environment, not an end in itself. This is because the effort involved in doing absolutely everything, and Acorn's relatively limited resources, limit what can actually be produced which means that their product range will continue to lag behind. Instead of trying to do that, Acorn should adopt widely-available core technology and build their personality on top of it. Of course, this does not mean that they should compromise on standards or quality, but rather they should take advantage of already-available technology and save their effort for what they can do best. What I propose is using a widely accepted, portable, open-source operating system (you probably know what I mean...) as the core of their system. This could run on either Acorn-produced hardware or the "industry standard", giving R***OS far more exposure and opportunity for wider use than if it were restricted to the Acorn platform only. The reason for abandoning both the requirement for proprietary hardware, and the RiscOS operating system core, is to free the skilled and resourceful people that Acorn possesses in order to allow them to work on improving the desktop environment, applications, programming tools and compatibility libraries. It will also allow them to dispose of what must be many thousands of lines of RiscOS system code, a work of genius when it was originally coded in assembler but which must by now be almost impossible to maintain and expand effectively. The hardware should consist of ============================== Acorn's primary (and indeed only) hardware product should be an StrongARM-based motherboard on the standard AT/ATX form factor with PCI expansion. There is no point in their engineers designing cases, power supplies or graphics adaptors: these will soak up much development investment for a very limited return, so just buy them on the open market instead. This motherboard could be sold alone for those enthusiasts who would like to build their own customised system, or complete systems could be assembled by dealers and sold to those who would prefer the assurance of having a completely built and tested product. The basic board should include an IDE interface, floppy interface, keyboard and mouse, multiple serial and parallel ports: this could all be incorporated at minimal cost by using current ARM components and standard PC peripheral chips. An added bonus would be a basic Ethernet, sound system and graphics, thus being a complete computer in itself but still having potential for expansion later. Naturally, the tradition of having the complete OS in ROM will have to be abandoned. There is simply too much effort and risk involved in producing, distributing and installing upgrades and new versions, which tends to slow down the pace of OS development - customers were kept waiting for RiscOS 3 for well over a year, even though its existence was an open secret and developers were allowed early access - and even longer than that for the FileCore upgrade. The entire concept of a floppy-only machine should also be abandoned (although the OS would support fully discless operation from a suitable server). Having this motherboard available gives the potential for various hardware configurations. At the low end, the single board in a case would provide a computer with graphics and networking available in a usable form at low cost. The top-of-the-range system could use the same basic part, but with the addition of full-spec SCSI, high speed Ethernet and accelerated graphics all on PCI cards. Anything in between is possible and workable. The software should consist of ============================== At the heart: a Linux (preferably) or BSD (if unavoidable) kernel, with drivers for all current RPC hardware and a selection of popular PCI cards. If a manufacturer won't provide sufficient information to allow a driver to be written, or will only allow it to be distributed under an NDA, then forget them: there are plenty of other expansion cards out there that could be used instead. Coming free with this will be the X Window System: all that needs to be done is to adapt it to give the look and feel of RiscOS. Some attempts are already being made to give X a friendly and consistent front end: CDE, KDE and Gnome are all good steps in that direction, but if necessary Acorn could develop their own window manager and libraries. The basic server and desktop will need extensions, either the current de facto standards (assuming that they are satisfactory) or developed (if none are already suitable) to provide drag-and-drop, file typing, printing and other RiscOS facilities. If possible, vendor extensions could be added to the X server or protocol in order to support Acorn fonts, Bezier paths and sprite images. There will have to be a desktop filer of course: not simply directory displays, but also with URL links (as in Windows 98 or KDE) and integrated with a Web browser. A major innovation would be image filing systems and transparent access to archive files, as provided by SparkFS (and preferably not the half-baked implementation that KDE has). BBC Basic (with assembler) will have to be there, of course. For the programmers and application developers, there should be C programming libraries giving source compatibility (as far as is possible) with existing Toolbox or RiscOSLib applications. The libraries could also provide other compatibility measures, including an emulation of the RiscOS file name format (i.e. UnixLib in reverse) to keep familiarity and ease the task of porting existing RiscOS applications. The standard RiscOS applications should be ported (and hopefully improved in the process!), and provided as a standard desktop set. There will obviously be a thinly disguised "Task Window" (a terminal shell by another name). For the RiscOS diehards, a CLI-workalike shell could be provided to again ease the transition and to allow existing Obey scripts to be run. Compatibility, both with older Acorn systems and with industry standard, is an essential part of any computer development. To this end, servers and applications should be provided to interface with existing applications and Acorn systems: for example, a Samba/NFS server to provide file and printing services to both Acorn and PC clients; a boot/file server for NCs; an Internet firewall providing a Web proxy and mail server. Apple connectivity would also be available to keep Xemplar happy. Open-source software to perform these tasks is available for anyone to download and use: however, such software can be difficult to install and configure correctly, especially for non-technical users. Having the same software provided installed and preconfigured as standard, with friendly GUI-based tools for setup and maintenance, would allow them to be effectively used with minimum support effort required (a major advantage for schools and other non-technical users). Naturally, Internet access (Ethernet, dialup or ISDN) would be available as standard. Acorn's income would come from selling ====================================== ...a variety of products, all using the same basic technology. Non-technical users, schools and anyone who just wants to get on with using their computer without having to worry about what is underneath would prefer to buy a complete system (hardware and software): built, installed and tested as a complete package either by Acorn or a dealer. They would also like to have technical and after-sales support, again provided either by Acorn or a dealer, either at additional cost or included in the price of the package. These users would have the assurance of knowing that their system is set up and working (or at least it would have been when delivered!) and is fully supported by the supplier. Those who would like to have Acorn hardware but who would not need the handholding as provided above could purchase an Acorn motherboard and processor, and assemble their own system from that along with standard PC hardware. Software installation (from a CD provided with the motherboard) would be up to the user, although again enterprising dealers could provide an installation and support service. It should also be possible to purchase the software on CD alone. This would allow those with existing Acorn hardware (of a suitable specification) to run R***OS: the current RiscOS in ROM would be relegated to a startup role only. The software should, of course, be provided in source form. Although this may be heretical to Acorn's current position, I believe that having Acorn as the only source of the operating system, standard applications and development tools has led to stagnation in development and restricted improvement and innovation. Providing the software in this form would also allow it to be integrated with existing kernels and libraries for other architectures, thus allowing it to run on other hardware platforms. Although again such an approach would not be officially supported by Acorn, it would allow their desktop environment to run on cheap hardware (far cheaper than anything that Acorn will ever be able to produce themselves) and will provide a route for newcomers to try out R***OS and experience the benefits without having to spend a substantial amount of money first. Development tools will be required, of course. The "official" product could be a development of Norcroft, or ARM's (rumoured) much-improved C/C++ compiler (currently only available on PC and Un*x). However, if providing either of these turns out to be impossible or too expensive, then GCC could be used instead. Another major advantage of this system is that it will not only be able to run the RiscOS desktop but also powerful open-source applications (e.g. Mozilla, XEmacs, Gimp). All of these are freely available for anyone to download and install; however, doing so is sometimes difficult for those without technical knowledge and sufficient Internet bandwidth. The non-enthusiasts could buy the software from Acorn, precompiled and configured all ready to install (assuming that their distribution licences permit so) and with technical and dealer support available. Enthusiasts and hackers will prefer to download and install the software themselves, with no support or help: again, there is a choice. Who it would appeal to ====================== Schools and novice users: those who want the easy-to-use Acorn system supported by dealers and a band of enthusiasts. They could buy PC hardware (to satisfy their budgets and the "industry standard" advocates), but retain the RiscOS desktop and their familiar applications. The same system could also run on their existing Acorn hardware (thus further demonstrating the long lifespan and cost advantages of Acorn technology as opposed to Micro$oft/Intel). Networked organisations could return to "100% Acorn" systems, servers as well as clients, rather than having to run NT to provide file/printer and Internet services. Current enthusiasts: those who enjoy using RiscOS and would like to keep supporting and using Acorn, but get depressed by the price differentials and lagging technology. Developers and users of future Acorn technology: embedded systems, portable appliances and network computers. The new Acorn system could be used for software and server application development, and also to provide network services. Again, such a solution could be provided by "100% Acorn" systems rather than using NT or third-party Un*x. Hackers and Internet fans: those who would like to use non-Micro$oft systems on principle (and be able to say "Windows, pah! Real hackers run Un*x...") but also have access to the latest networking technology and powerful applications. Those who would like to use alternative hardware. Support for the Intel architecture - even if not officially supported or even sanctioned by Acorn - will allow R***OS to run on exotic hardware (e.g. industrial/single board computers, portables, multiprocessor systems) which would be totally uneconomic for Acorn to develop or produce themselves. Having the software available in source form and based on a portable kernel would also allow the possibility of having R***OS running on Alpha, Sparc or future Intel platforms. Who it won't appeal to ====================== Writers and users of Zap, Impression and other RiscOS applications which couldn't be ported: unfortunately the development of these would reach a dead end (if not having already done so). However I have no doubt that Ovation Pro and similar applications could easily be adapted to the new system, and the Zap hackers and extenders could write for Emacs instead... Acorn peripheral builders: the proprietary expansion architecture would have to disappear. However, designing for PCI instead would give them access to a potentially far larger market: unfortunately, one in which they may also sink without trace. Bill Gates (and his representatives on Earth). A system with the friendliness of RiscOS and the appealing desktop of Windows, combined with the power and versatility of Un*x, but cheaper and with less overhead than NT. Still with the familiar Acorn name, and able to take advantage of their established position and experience in the UK education and enthusiast markets - but without the premium price of proprietary hardware which is forcing many individuals and educational customers on to PCs. Will it ever happen? ==================== Probably not. But I can hope... [P.S. OK, so that was a bit long. If you want to reply then *don't* quote the whole of this message! Followups to c.s.a.advocacy only.] -- Jonathan Marten work: jonathan.marten@uk.sun.com http://www.keelhaul.demon.co.uk play: jjm@keelhaul.demon.co.uk