Original URL: https://www.theregister.com/1998/09/30/how_to_drop_dos/

How to Drop DOS (and still play games)

Emulators are the key, a reader writes

By Paulius Stepanas

Posted in On-Prem, 30th September 1998 15:33 GMT

Recently, a number of big players have started exploring how they might drop legacy support from their new systems. Examples include Windows NT, Merced, and now Transmeta. Technologically, this is a good move, as it makes it possible to discard the 20-year-old x86 architecture and all its bizarre quirks. It also allows potential independence from legacy DOS and Windows APIs. But, there is a problem: how to guarantee a smooth transition for current computing environments to a completely incompatible system? And just how important is this? It is critical to maintain backwards compatibility, but with one major proviso: DOS/Win95 compatibility becomes irrelevant in the presence of a complete DOS/Win95 emulator. Before exploring the solution, let's examine the problem. Computer gaming is arguably the major driver for consumer upgrades of both hardware and OS. This is illustrated by recent successes with short-lifecycle, 3D graphics cards, where gamers have been persuaded to update their graphics capabilities (and often their CPU) two or three times in the last two years. Computer gamers are also most resistant to changes in the support of legacy software, and with the most reason. Where large corporations continue to use old desktop operating systems purely on the basis of monetary concerns, for gamers migration to an incompatible system makes it impossible to play their old favourites, for which they still remember paying hard-earned money. The corporation will, at some point, upgrade their software to the latest and greatest. Gamers will never have this option, as a game written for DOS and a 386 will never be updated to Windows 2000. This would appear to be an irreconcilable impasse. But examination of certain communities on the Internet suggests a viable compromise. In a word, the answer is an emulator. On the Internet, there is a thriving community of emulator users. Complete, cycle-accurate emulators are available for the Commodore 64 and Apple II, amongst others. These run at reasonable speeds on 486-generation computers (some better than others). They are programmed by hobbyists, often individuals with no support beyond the emulator community. At a slightly different level, emulators of Windows 3.1 are available that run on UNIX-based systems. Here, the complexity has greatly increased, but the level of emulation has also changed, since emulation of hardware is no longer required. If these emulators can be written in people's spare time, a large corporation should have no trouble writing an emulator for DOS or Windows 95. But it still remains to examine the viability of an emulator, so let's examine the DOS situation. The last games written for DOS alone mostly ran fine on a Pentium 60 (in fact very few needed more than a 486DX2/66). The standard, high-end, consumer CPU is currently a Pentium II/350 (roughly). This runs approximately 24 times faster than a Pentium 60 (I'm guessing x4 for two generations, x6 for clock speed -- bus speed is irrelevant, as nothing extra is happening on the bus). Hence, it shouldn't be difficult to build a complete, cycle-accurate, DOS emulator that runs on Windows NT at faster than 1/24 the speed of native DOS on the same machine. All we need is for someone to build it. (And Microsoft is best placed to do so, but how does one let them know?) If you add a year to the equation, the same can be done for Win95. Naturally, one could argue that an emulator is unnecessary, as you can simply install a second operating system. Even ignoring the issue of hardware dropping legacy support (eg: Merced), this is not a truly acceptable solution. The goal is to make it unnecessary to reboot the machine (since we all know how annoying and lengthy a process that can be). Rather, we want to be able to store all our old games on the same system and run them at the click of a button. Of course, given the probable processing demands of the emulator, it would help if most other operations of the OS could be suspended during the operations of the game (or other applications). Gamers tend to be early-adopters for new releases of hardware and software, yet they will often play old favourites from their youth(when 8-bit computers were all the rage). If Microsoft, Intel and other companies expect this important market to upgrade to incompatible systems in the near future, they will have to make provision for the legacy requirements of these users. It took a while for Microsoft and Intel to realise the size and strength of this market segment. Perhaps it's time for them to re-examine their legacy-support strategies -- if they wish to stay on top of the game.® Click here for more stories