nagy%43198.hepnet@LBL.ARPA.UUCP (04/06/87)
Well, I figure the BIG feature for VMS V5.0 is Symmetric Multiprocessing. This is going to involve major internal changes in VMS. Currently VMS uses processor IPL level for internal synchronization and access to common data structures; this does not work in a multiprocessor environment except in the master-slave style currently used by VMS. SMP will require the use of mutexes and/or semaphores to provide the necessary interlocks. This is all likely to break privileged code many of us (myself included) have written to dip into VMS or to provide extensions to VMS. I guess I sort of accept this; some of my privileged code has survived 2 major VMS upgrades (V2->V3->V4) without requiring changes (other than reassembly and relinking). I think THAT is really amazing. Some of this was discussed at the previous 2 Symposia in the vein of "you better watch out...". I did gather that VMS Development expects NOT to have to require extensive (if in fact any) changes to device drivers - which are probably the most common examples of user-written privileged code. I'm not at all sure how they are going to pull that one off... Also, it has been mentioned that BACKUP will be undergoing a complete rewrite for VMS V5 with emphasis on performance but without sacrificing any of the redundancy currently provided. About time. Finally, I have heard the rumor that VMS V5 is actually RUNNING in the lab at Digital. Maybe V5 is sooner in our futures than we expected!