# DOS emulation

Contents

In the course of understanding various emulators and not-quite-emulators of legacy software, it is helpful to consider two very different ways of substituting Microsoft platforms over the past four decades: the DOSBox and WINE projects.

# DOSBox

Only one of these, DOSBox, is technically an emulator, though it is an unusual one. DOSBox emulates a set of hardware configurations that most of its users would rather not think about -- mostly a set of common PC platform targets from the 80s and 90s which have only subtle differences from one another -- in order to provide an environment which is compatible with most DOS software. Although all of these DOS platforms were/are based on x86 PC architectures and could theoretically be virtualized on most modern computers rather than emulated, the increased portability (e.g. to ARM platforms) of a fully emulated architecture, as well as the relatively low needs of the slower, older hardware being emulated, probably minimize the incentive to maintain a virtualization solution, and no one currently does so.

The emulator also provides the DOS environment itself. This is somewhat atypical, owing to the various DOS APIs being relatively easy to reimplement without copyright infringement. The FreeDOS project was the first to do this in 1998, after Microsoft themselves stopped supporting any official releases of DOS and otherwise requiring users to run Windows to access the large library of DOS software via its compatibility layer, and is fairly reminiscent of the more recent Google LLC v. Oracle America, Inc. fair use arbitration in this way. FreeDOS was intended to run directly on baremetal desktop computers of the time which probably would not have been powerful enough for emulation of anything more than a 286 (where most popular DOS software was designed for 386, 486, and Pentium machines); DOSBox does not include FreeDOS itself, but borrows a similar approach, and both projects continue to be well-maintained.

Managing various audio and video plugins to control output is usually one of the more unpleasant aspects of hobbyist emulation, and DOS has the distinction of having made this fairly miserable when it was contemporary as well. The preconfigured DOSBox target machine settings are helpful here, but emulating DOS still entails working with a DOS filesystem and an ancient command prompt. It has been said that where Apple platforms are a delight to use up until you attempt to run a game on them, Microsoft platforms are the opposite, and true to form, most popular uses of DOSBox do not involve spending much time in DOS itself. Many resellers of legacy PC games from the period actually ship their titles in prepackaged configurations of DOSBox designed to run easily in modern environments; as in so many emulation use cases, the game is the point, rather than the DOS platform itself.

# Wine

WINE, on the other hand, is not an emulator -- that's what the acronym stands for. Rather, WINE is a runtime compatibility layer for running Windows software on non-Windows platforms that replaces various Windows API calls and low-level libraries with their equivalents on other platforms. It is exceptionally performant, generally able to translate DirectX rendering calls in real time to cross-platform alternatives (such as OpenGL or Vulkan) with about one CPU thread's worth of overhead. Its compatibility with decades of Windows software is tracked at winehq.org, and it can be a very convenient alternative to needing an entire Windows virtual machine (and associated licensing) to run specific legacy applications.

However: it can be quite fussy when running complex applications. Wine defaults to storing various environment settings and copies of Windows libraries in per-app sandboxes, similar to how pyenv works for Python programming, called "Wine prefixes," and juggling these requires some command line expertise. It also tends to run a half-decade to a decade behind newer Windows frameworks (similar to how GIMP, Inkscape, et al tracked the Adobe suite's feature set in the 2000s); using .NET software wasn't trivial until fairly recently, and the newer Windows .msi packaging format can still cause headaches.

If there is a reason that WINE has never become more popular in preservation circles, it is probably because it solves the wrong problem in the wrong contexts. Its relative instability and fussiness generally makes it better suited to one-off solutions and individual needs to use a single piece of software; however, its primary use case of running software in an environment where Windows is not an option typically describes only servers[1], where there is a preference for running software that is more hardened and more reliable. This is unfortunate, because the WINE project has generated an enormous amount of documentation and technically liberated an enormous amount of software from a platform controlled by a single commercial entity; the WINE wiki is often said to be a better resource for legacy Windows APIs than Microsoft's own extant documentation. But the conditions under which an user cannot or will not run Windows for reasons that are technical in nature rather than ideological or circumstantial are nevertheless rare. However robust WINE itself may be, the ever-expanding set of Windows software that it is obliged to add compatibility for will likely always add uncertainty in practice.

DOSBox and Wine both have the advantage of being fairly monolithic projects with many contributors and no real competition where faithfulness of emulation is concerned. As such, they are often used as the basis for derivative projects. For example, https://github.com/danoon2/Boxedwine is a project that emulates Windows by emulating x86 Linux and running Wine on the emulated Linux.

[1]: Windows servers of course exist and are popular in enterprise IT contexts; they may as well not exist for the purposes of software preservation and access, as licensing and deployment are dramatically more prohibitive than with virtually all other tools available to curators, and there is a justifiably absolute preference for open source tooling of digital preservation platforms themselves, even when the platforms being preserved are otherwise encumbered.