HP Discontinue OpenVMS 238
simpz writes "The register is reporting that 'the ancient but trustworthy server operating system' OpenVMS has been discontinued. From the article: 'HP never really promoted its acquisition and OpenVMS suffered from a lack of development compared to HP-UX, itself suffering from competition from Linux. It was only a matter of time, but it's a sad end. Many of its old-time fans, your correspondent included, cherished a hope HP would move it to x86-64 – but since development moved to India in 2009, OpenVMS has been living on borrowed time. Now, it's run out.'"
Never hacked? (Score:4, Interesting)
Last time I heard VMS had never been hacked. Is that still the case?
It was the best OS I ever worked with. It'd be nice if they open sourced it.
RIP VMS (Score:5, Interesting)
There were few operating systems that handled loose-clustered networking as elegantly as VMS. Want to centralize user credentials? Easy, just place SYSUAF.DAT on a shared volume. And since the files could have structure, you could lock individual user records for editing rather than the whole file.
Another great feature was the concept of "quorum". Quorum, as in the organizational term of the number of people present at a meeting necessary for it to be an official meeting of an organization, was the number of reachable hosts necessary to conduct business. Say you had a redundant banking site - and the link between them would go down. If they are a redundant configuration, they would continue to process transactions - with their database quickly diverging. Using quorum nodes, you could set up three hosts on three sites - two major server setups and a simple workstation somewhere central - and voila, no single point of failure.
Besides, there is a magnificent book, "OpenVMS Internals and Data Structures", which so elegantly and wonderfully describes operating system design.
I really, really hope that OpenVMS could be open-sourced and this codebase might serve as the base for a community-written x86 port.
Re:When will it be open-sourced? (Score:4, Interesting)
However, the real beauty of VMS wasn't so much it's architecture (though that had a lot of good points) but the incredible quality of DEC's implementation. Bugs were for the competition.
While I used VMS extensively and liked it in many ways, this is just silly.
When VMS 4.0 was released (the first version to include DCL command line editing), we had some unexplained crashes in our cluster. We eventually tracked it down to a bug in the command line editor (yes, it ran at least partially in kernel space). We had a local "competition" to see who could find the shortest number of keystrokes that would crash the system. The winner: 4. Yes, you could crash VMS 4.0 by getting an unprivileged command prompt and typing 4 characters (didn't even need to hit RETURN).
The bug was fixed in 4.1.
VMS was doomed when HP bought it (Score:5, Interesting)
When the amount of development your OS gets suffers "compared to HP-UX" you are in astonishingly deep trouble. I have had three run-ins with HP-UX, first in 1998, next in 2004, and finally in 2010 (when my current job retired all it's existing HP servers and moved to Solaris). When I encountered HP-UX the first time, in 1998, it seemed to be at least 10 years behind the times. Very little had changed in 2004, which meant that it was falling farther and farther behind each year. In 2010 it seemed little better than it had been in 2004, and I guess that management agreed, since we finally cut the cord and moved on to something that was, at least by comparison, more up to date.
I also used OpenVMS in the early 2000s, and it was capable, but idiosyncratic (record structured files were a PITA, and the file versioning was no replacement for proper version control. I really liked logical names, however, and the global symbol table was useful). It had a head start on lots of other OS's with respect to clustering features (cluster wide file system, message queues, and distributed lock management was all built-in), but much of the userland was GNU stuff ported over on the POSIX layer. DEC seemed to have given up on the whole "innovation" thing and was just milking existing big contracts.
Re:When will it be open-sourced? (Score:4, Interesting)
Re:Never hacked? (Score:4, Interesting)
uhhh. no.
Mitnick social engineered his way into VMS. he did not "hack in". He used the telephone and convinced a flunky to start a command interpreter on the modem line he was dialed into. Clever? yes. A skilled hack? only of humans.
Re:When will it be open-sourced? (Score:4, Interesting)
DCL didn't run in kernel space, it ran as supervisor code (the four levels were user, supervisor, exec, and kernel). DCL sat above the stack in the user's address space (the user had two address spaces) so when it ran a command the command code was loaded into the regular user heap and executed without starting a new process. The command would just "return" at the end and you'd be back to the command interpreter.
Anyway, if you could crash the whole system with DCL the problem was likely in QIO, not in DCL.
Re:When will it be open-sourced? (Score:3, Interesting)
Re:When will it be open-sourced? (Score:4, Interesting)
The real big difference I felt between Unix and VMS was the orientation. VMS was fully intended to be what we'd today call an enterprise system. It was for corporate office to run as a server, for database management, for batch processing, etc. Unix was oriented towards small departmental computing. Late 80s had Unix growing up a bit more but it still had a much looser feel to it whereas VMS felt like you needed a suit and tie. At that time too Unix was pretty efficient, it really depended on what you were doing though; lots of users or heavy duty I/O and VMS tended to win, whereas few users and Unix felt more responsive. Unix was also always more open; cheaper, more third party applications, free development tools, etc. It changed in early 90s though when Unix got that corporate feel and all the big players wanted a piece of the pie and started splitting into factions.