On 7/27/07, Joerg Schilling <[EMAIL PROTECTED]> wrote: > > "Brian Gupta" <[EMAIL PROTECTED]> wrote: > > > > using Jaluna code plus OpenSolaris code (for an approximation of SVR4 > e.g. > > > filesystem and other code) as a starting point; seems to me that > > > Chorus/MiX would have been on a level of coolness approximating that > > > of HURD, Apollo Domain/OS (had it ever been open sourced), Spring > (ditto), > > > or the like. > > > > > > Of course, some would say that the microkernel is semi-dead, given the > > > message passing overhead (or whatever the fully clueful version of > that > > > argument is). But the level of flexibility, extensibility, and in > some > > > multi-server implementations, ability to recover from even driver or > > > filesystem errors without crashing, IMO remains interesting. > > > > > > Microkernel design philosophy is resurrecting itself in surprising > places. > > Take a look at the work the Linux folk are doing to implement ZFS in > > userspace. > > Please do not confuse microkernel with multi-context kernels. > > Solaris _is_ a microkernel architecture and not a monolith. > Solaris does however run al the code in a single kernel context.
Could you please explain how you see Solaris's kernel being a microkernel? (Are you thinking because Solaris supports dynamically loadable kernel modules that it is a microkernel?) I would guess that in future some of the functionality will be in userland > even on a traditional single-context kernel like Solaris. This is already > true for CD/DVD/BluRay writing (see cdrecord and the SCSI Pass through > driver that I invented in August 1986), it is also true for libusb and > the future Bluetooth stack. What do you mean by context in this context? (Sorry I really should correct the wording, but it gave me a chuckle). Jörg Cheers, Brian
_______________________________________________ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
