I am always interested is watching and reading discussions around what Open Source or Open Standards exactly are. Some folks are very passionate that if a particular thing: Object, projects, whatever, does not meet their exact definition of open, then it is therefore not. Not good. Not Open. Not open. It is bad.

I was presenting at the Z Series conference 3 or 4 years ago, and during my talk about VM I mentioned the the OS was "open source". In the Q&A afterwords I was challenged on that characterization. Vm was not GPL. It was not Apache. It was not any "Open Source" license at all. It was and is a copyrighted work of IBM. It was also the first real Virtualization OS, but I digress.

In the early days IBM had no real use for VM other than as a migration tool. As such, they had an option where you could order the source code. It would show up on the tape, and one could load it to disk. In fact, to put on a patch required have the source code because the patch updated the source, and then rebuilt the module or nucleus or whatever.

The source code was open in the sense that it was there for all the world to see. And having access to it sparked all sorts of innovation. One example was the V/ line of products from then VMSI. VM had a problem. Several of them in fact. One of them was that IBM did not love it. It was always a step child to MVS. MVS sold hardware: it took 10 times as many resources to log in to TSO (Time Sharing Option) as a user than to VM as a CMS user. CMS was the end user interface bit to VM. The part that defined the editor and the file systems and so forth.

Since VM did not get a lot of development cycles from IBM, it did some fairly stupid things sometimes. One was that when it came across an internal error, it would abend. Even stupid little things would cause a complete system reboot. VMSI created a product that intercepted the abend, backed it up a step, then made an attempt to do something to fix the problem rather than just break. In most cases, VM could stay up, or at least limp along with only a partial disability until an outage could be scheduled.

Even better, they had another product that would take a dump of the OS before it started fixing things so that the problem could be looked at and shot while the system stayed up. There was even the option of a snap dump so that the whole memory core did not need to be dumped, just the relevant bits.

IBM added features like these to later versions of VM once they saw the wisdom of them. And of VM for that matter.

I would love to have features like this in Linux or BSD today. They came about because VMSI had access to the source code, even though VM was not GPL or whatever.

Amdahl was another company that benefited greatly by access to the source code to VM. VM was a virtualizer... and it virtualized the mainframe. Amdahl wanted to build Plug Compatible Mainframes, but the book the IBM had published, the 370 Principals of Operations or the "POO" was not actually 100% accurate. VM gave Amdahl all sorts of ideas. They created the predecessor to PMA called VMPE (VM Performance Enhancement) so that some virtual machines could get a little extra boost. PMA (Preferred Machine Assist) was IBM's late to market answer to Amdahl's innovation. Amdahl then created its tour de force product: MDF, or Multiple Domain Facility. Again IBM had to respond and created PR/SM. PR/SM (pronounced "Prism" usually) was... VM in the microcode of the mainframe. No. Really. IBM has always denied this, but I watched some CE's at the CE console once, and saw the error messages: that was VM in there. It was years before PR/SM could do what MDF did. I knew several folks at IBM back then that were deeply worried that folks would figure out just how good MDF was before IBM had something to compete with it.

Note: Updated 1/21/2008 to correct a technical point: Originally I wrote the IBM's answer to MDF was LPAR. It was in fact called PR/SM, and the "Domains" of MDF were called "LPARS" or "Logical PARtions". Thanks to Richard Meyer for catching that faux paux

One last point here: The best professional training I ever went to was Amdahl's VM diagnostics class. That class was written because they could look at the source and develop material about how to read dumps, and had the ultimate reference: they could show how the thing actually really worked!

Having open source... access to the source code, lead to these things, and they ultimately benefitted the customer.

It was not Open. It was just open.

Hey Steve.. is there a point in there? Only that openness comes in many flavors, and honest attempts to be open should not be greeted with derision and scorn if they are not 100% in line with a particular definition of open.

OOXML and the like can be as derided as one likes though. While I am for folks (I include companies in that) trying to be more open, It really gets up my nose when they start trying to use openness as a trap. I think there are several conversations about patent traps elsewhere on the forum recently: I am so not talking about that kind of "open".