Re: RFC: adding fd-passing to win32
Lawrence D'Oliveiro <ldo-kC51rjc0cqs1dNLGKZg6taU/[email protected]> Mon, 22 Aug 2022 22:48:29 +1200
| Newsgroups | gmane.comp.freedesktop.dbus |
|---|---|
| Organization | Geek Central |
| Message-ID | <[email protected]> |
On Mon, 22 Aug 2022 09:05:40 +0200, David S. wrote: > My understanding of Cygwin's approach, which is to provide a more or > less POSIX compliant API to Unix based software to compile and build > on Windows. Precisely. And POSIX is at the core of Free Desktop efforts. If your platform doesn’t support POSIX, I for one cannot see many open-source developers bothering to make the extra effort necessary to support the peculiarities of your platform. > Even though D-Bus itself started off in the Unix-like landscapes, I > would find it valuable if D-Bus was more Windows native when used on > Windows. I don’t know why you even need D-Bus on Windows. Microsoft was at the forefront of much more elaborate high-level IPC efforts, dating back to COM/DCOM, OLE and even CORBA from the 1990s. D-Bus is a much more limited concept, considered to be just about sufficient for Free Desktop efforts, and not much more than that. In other words, don’t you already have much more powerful capabilities, supported by the platform vendor itself? > That removes the hurdle of needing Windows applications > wanting to use D-Bus to also need to pull in Cygwin as part of their > build stack - at least when needing specific D-Bus features such as > the FD passing. That seems to be different beasts on Unix vs Windows. That is a fundamental limitation of Windows, which Cygwin manages to paper over. Since Cygwin gets rid of the extra effort necessary to support the peculiarities of your platform, it seems like a win-win scenario to take advantage of that. Unless and until Microsoft itself comes up with a better solution ... > So if D-Bus on Windows can be used more natively, that makes it more > likely to grow further on the Windows platform as well. Porting an open-source project to Windows will increase the sheer number of users, but it will not proportionately increase the pool of contributors back to the project. In other words, it will increase the pain, but not necessarily the gain.