Cygwin: A Free Win32 Porting Layer for UNIX Applications

9. Proprietary Alternatives

When we started developing Cygwin, alternatives to writing our own library either did not exist or were not mature enough for our purposes. Today, we know of three proprietary alternatives to Cygwin: UWIN from AT&T Laboratories, NuTCracker from DataFocus, and OpenNT from Softway Systems.

UWIN ("UNIX for Windows") was developed by David Korn for AT&T Laboratories. Its architecture and API appears to be quite similar to our library. Its single biggest advantage over Cygwin is probably its more complete support for the UNIX security model. UWIN binaries are available for free non-commercial use, but its source code is not available.

NuTCracker, by DataFocus, is another proprietary product that is built on top of the Win32 subsystem. Version 4.0 of the product appears to be quite complete, including such features as support for POSIX threads.

OpenNT from Softway Systems takes a markedly different approach by providing a capable POSIX subsystem for Windows NT, implemented with the Windows NT source code close at hand. At least in principle, writing a separate POSIX subsystem should result in better performance because of the lack of overhead imposed when implementing a library on top of the Win32 subsystem. More importantly, by avoiding the compromises inherent in supporting both Win32 and POSIX calls in one application, it should be possible for OpenNT to conform more strictly to the relevant standards.

However, there are two substantial drawbacks to OpenNT's approach. The first is that it is not possible to mix UNIX and Win32 API calls in one application, a feature that is highly desirable if you are attempting to do a full native Win32 port of a UNIX program gradually, one module at a time. The second drawback is that OpenNT does not and cannot support Windows 95/98, a requirement for many applications, including the GNUPro development tools.

The lack of source code, coupled with the licensing fees associated with each of these commercial offerings, might still have required us to have written our own library if we were faced with the same porting challenge today.