David L. Craig wrote:
On 14Nov28:1259-0500, Marty wrote:We have to get rid of udev and dbus together, because they will be merged in kdbus. My problem is dbus because it goes everywhere and I wouldn't know where to start. From a systems integration standpoint my top questions are: - Is it possible to start replacing dbus with 9p, starting with eudev, then replace it in forks of the other big targets like login, policykit and PAM? (Do the GUI parts last, maybe using a shim.) - Can we develop a strategy to complete all this in a release cycle plus 1 year (maybe 4 years max?) - What is the status of 9p support in Linux? (9P_NET and 9P_FS) How do they compare with 9p2000 and are they ready for prime time?
I'm not sure that kdbus is that much of an issue, for a few reasons:- it seems to add capabilities that are missing from the kernel (at least there's been discussion of a faster IPC mechanism in the kernel community for years) - the kernel folks (and Linux in particular) are a lot more hard-assed when it comes to letting stuff into the kernel that's not ready for prime time (useful, tested, documented, supported, doesn't break userland, etc.)
If kdbus gets to the point of actually making the cut for inclusion into the kernel, that might actually solve some problems associated with having it outside the kernel with less stringent QC.
On the other hand, we would then have one more thing in Linux that breaks cross-platform compatibility with the BSD and illumos worlds.
Miles Fidelman -- In theory, there is no difference between theory and practice. In practice, there is. .... Yogi Berra
