From: Jim S. <ja...@ne...> - 2004-06-23 14:44:21
|
eg wrote: >> >> That's pretty discouraging. I don't quite understand how someone >> could participate in a standards process without having the foggiest >> idea of how other systems have solve the problem. It's not like the >> VMS lock manager is a secret or isn't generally considered the gold >> standard, and has support lock conversion from the beginning. His >> arguments are utterly lame. A downgrade is without contraversy. An >> upgrade is deadlock prone, but that's an application problem, not a >> lock manager problem (except to report it). >> > > FYI: In this case, the person in question should have some idea of > "how other systems have solved the problem". It seems he has been > working on thread libraries at DEC/Compaq/HP for some time according > to his web page: http://homepage.mac.com/dbutenhof/dave.html I don't know him so I can't really say. But operating system types tend to extremely insular. When I was at DEC, for example, almost none of the people working on VMS had experience on operating systems other than RSX-11M or M++ (Cutler came in the box when DEC bought -11D). The scheduler was an anomoly have been written from Dick Husvedt who had worked the XDS 940. Not a one had been exposed to Unix. On the other side of the fence (in a different building, generally) nobody in the Ultrix group had the slightest experience on VMS. A VMS guy did wander over the Ultrix to do an experimental Ultrix driver for the VMS ODS-2 file system, but then the VMS group wouldn't take him back. Cross fertilization in operating systems is extremely rare. That said, it's clear that his job was to allow Posix conforming programs run on VMS, not to allow VMS programs to port to Posix. DEC/Compaq/HP have never been in the business of building life boats. Not that I would be so cynical as to suggest that it was goal to ensure that Posix threads were semantically incompatible with native VMS software... -- Jim Starkey Netfrastructure, Inc. 978 526-1376 |