home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!cs.utexas.edu!sun-barr!news2me.ebay.sun.com!seven-up.East.Sun.COM!tyger.Eng.Sun.COM!geoff
- From: geoff@tyger.Eng.Sun.COM (Geoff Arnold @ Sun BOS - R.H. coast near the top)
- Newsgroups: comp.protocols.nfs
- Subject: Re: lockd
- Date: 23 Jul 1992 17:52:08 GMT
- Organization: SunSelect
- Lines: 63
- Message-ID: <14mrk9INNis@seven-up.East.Sun.COM>
- References: <1992Jul21.174721.26888@nntpd.lkg.dec.com> <14hpi3INNppo@seven-up.East.Sun.COM> <nje1jic@rhyolite.wpd.sgi.com>
- NNTP-Posting-Host: tyger.east.sun.com
-
- Quoth vjs@rhyolite.wpd.sgi.com (Vernon Schryver) (in <nje1jic@rhyolite.wpd.sgi.com>):
- #In article <14hpi3INNppo@seven-up.East.Sun.COM>, geoff@tyger.Eng.Sun.COM (Geoff Arnold @ Sun BOS - R.H. coast near the top) writes:
- #> ...
- #> Why? Because back in 1986 it wasn't practical to create a multithreaded
- #> lock manager.
- #
- #
- #This is the first even remotely official statement I've heard that any
- #of the various versions of lockd owe much to the simple locking demo
- #program shown at USENIX or whatever in Feb (I think) 1986. By
- #listening to what was said in meetings of B.Lyons' group at the time, I
- #inferred that it was not intended to be a product. Are you referring
- #to the same demo?
-
- OK, let me dig out my notebooks. 1986 was imprecise: I should
- have said 1985-86.
-
- I first got involved in the lock manager during the latter part of
- 1985, in email discussions with Jo-Mei Chang about the state of the
- work that she was doing with Bob Lyons. That was the project that led
- to the lock manager that was demonstrated at Uniforum in Feb of 1986 (I
- think - that was also the first demo of PC-NFS, and I was pretty busy
- at the show). That code cetainly formed the basis for the first FCS
- product (according to the SCCS history).
-
- In the summer of 1986 I took on temporary ownership of the lock manager
- source in order to add the file sharing and other pieces needed for PC
- clients. I also cleaned up some areas of inefficiency, but I didn't
- change any of the fundamental algorithms. I then passed the token back
- to the West Coast group. The subsequent history is murky, and includes
- a fair amount of interaction with the folks at Summit...
-
- #It was not <<technically>> "impractical to create a multithreaded lock
- #manager" in 1986. Perhaps no one had the time or inclination, but it
- #was certainly technically feasible. "Multi-processing", "threads",
- #and so on have not changed much since 1986.
-
- It is true that back in 1985-86 there were a few examples of LWP-style
- multithread packages. What were missing, of course, were the fully MT-safe
- runtime libraries (everything from stdio to rpc) and MT kernels
- to support them.
-
- # It is true that more
- #companies are shipping multi-processor UNIX systems today, but that is
- #not relevant. You don't need more than one processor for a
- #multi-threaded program.
-
- Natch.
-
- # You don't need klunky things like System V
- #shared memory and semaphores, which were around in 1986 and might have
- #been in SunOS--I don't remember. Lockd on SunOs of that vintage could
- #probably have used vfork(), although that would have been somewhat kludgy.
-
- I think that the first lockd work preceded even the SunOS 4.0 memory
- rearchitecture. Prior to that, vfork would have been a killer.
-
- Geoff
- --
- ---------------------------------------+--------------------------------------
- Geoff Arnold, PC-NFS architect | Thought for the day: If Unix had been
- (geoff.arnold@East.Sun.COM) | developed in England, we'd all be using
- SunSelect, a Sun Microsystems Business | BCPL (and migrating to BCPL++)...
-