Posted by: Mark Fontecchio
mainframe open source
We received some letters from readers this week in response to Robert Crawford’s column on making the case for IBM to open source the mainframe. Here are two of them.
I have 45 years in mainframes (1964 to present) and worked for IBM for 30 of them. IBM will never open source Z/OS, Z/VM, or any of its other flagship products. I used IBM’s open source code in the 60′s and remember clearly when OCO was announced. You write of the possibility of essentially customer “innovation” but innovation implies “hooks” into the nucleus and other key parts of the system loaded in the link pack areas. Yes, those hooks work on day 1, but on day 2, when IBM upgrades the source, maintenance becomes a nightmare, so much so that some customers refused to upgrade for years! Additionally, when things went awry, IBM was always blamed first – with demands that they find the bug ASAP – at no charge of course. No blame ever accrued to the purveyor of the modified hook. IBM was blamed for making the system “too complicated.”
Then there was the outright theft of intellectual property i.e., the code in those flagship products. I can write long lists of what open systems don’t have, even after 15 years of “development.” Open system vendors like to use words like “virtualization” and “data sharing” that sound like they are doing the same thing as a mainframe – but when pushed they are forced to admit they are not “really” doing what those words imply. They have had at least 15 years to copy IBM, but so far no one has come close to the reliability and availability of mainframes which have implemented true virtualization and data sharing for some 30 years now.
Don’t forget the mainframe hardware is tuned to mainframe software and vice-versa. The investment in that technology is now in the trillions of dollars – and I believe has the most patents for both hardware and software innovation in the world.
IBM is not going to give that away again!
I think you let IBM off the hook too easily. I have been in the business for over 40 years as a systems programmer.
First off I was a project manager at GUIDE for about 5 years when this was announced (IBM going OCO). There was a huge swelling of anti-IBM feelings at this point. The ceiling was raised with people criticizing IBM for its position. IBM finely responded that when OCO happened IBM promised to make documentation so good that you shouldn’t need source *AND* that all programming interfaces would be fully documented.
That sort of allayed the criticism and resentment subsided to background distrust. At first IBM seemed to honor their promises with better documentation as it did seem to happen.
Then after about 5 years the opposite started to happen. What was freely available in information was cut back to the bone.
After about 10 years IBM promises of better documentation went up in smoke. Case in point IBM’s COBOL (new) compiler spit out messages that were *NOT* documented. IBM never put out a messages and codes for the damn thing. I had a field day of making IBM look bad because of this. Then the ultimate chutzpa: IBM came back and said the messages were self-documenting. Hahaha. Then IBM started to make waves and actually charging users for documenting interfaces for IBM’s routines. I have heard $100K cost for a peek at a document which should have been open to everyone that bought the manual.
That was one case now the next is a classic. IBM wanted $150K for documentation so the user could write his product to sell to the “public”. IBM sold it to him and then about 2 years later they shipped a fix out which changed the interface and did not tell the end user that what was happening. This is easily done in SMPE with hold data. The vendor got a black eye because IBM did not tell people this fix would make a product stop running. I have heard that this product is European based but has some users in the US as well.
IBM charging for looking at documentation that should be available (at nominal charge ie cost) is NOT even close to coming close to the promise kept many years ago. The newer people in the business do not have a clue as to how important source is. Example: 30 years ago there was a utility that was part of Logrec, IBM’s facility to report hardware & software errors. This utility to run the report was extremely expensive CPU wise to run (we were against the wall ie largest CPU available and fastest and we were at 100 percent 7X24). I went to the fiche and after a few minutes of poking around I found why the damn thing was CPU intensive. There was a (4K?) table that when it got a record to update the table it went through sequentially and it took forever. I sat down and thought about it and figured out how to cut down the run time to 30 seconds or less just by indexing into the table and adding to the field directly. I opened an APAR and sent the documentation in with my suggested fix. IBM refused to implement it. What I did was to replace the IBM module in question with my own.
I know this is all water underneath the bridge but people have let IBM off the hook too easily nowadays. It’s time to hold their feet to the fire and hold them to their promises that were made when they first announced OCO.
- Ed Gould