Haha, good times: Larry Ellison regales noobs about when Oracle staff almost didn't get paid

Re: Sage advice

Agreed.

He got very very lucky once.

Then brutal time and time after again.

Re: Sage advice

Just because he foresaw the usefulness of a transaction processing database when IBM didn't doesn't mean he's all seeing.

Ellison's foresight is somewhat exaggerated in any case.

Oracle is widely credited with having the first commercial RDBMS product, shipping the first Oracle product in 1979, if Wikipedia is accurate.

However, IBM was only two years later with its first commercial RDBMS, SQL/DS for DOS/VSE. And DB2 appeared only two more years after that.

More importantly, while IBM's System R was never properly made a product, it was used at some customer sites. It was installed at Pratt & Whitney in 1977, two years before Oracle had a customer.

And the canard that "IBM didn't recognize the value of relational databases" is unfounded, too. Certainly Codd and his team understood why the relational model was important; that's how System R and SQL got developed in the first place. They were part of IBM. IBM's a big organization - there isn't a singular "IBM opinion" on every subject.

And even if some in IBM management recognized the commercial potential of relational databases, they had to make a case for them, taking into account the value of IBM's non-relational database business. SYSTEM 2000 (a hash-based database, something of a forerunner to today's KVS NoSQL DBMSes) and IMS DC (a hierarchical database) were significant IBM products, and IBM also made a lot of money from mainframe customers using ISAM and VSAM indexed files in lieu of DBMSes and paying for storange and CPU for them. Dedicating development, support, and marketing resources to Yet Another sort of DBMS was not an easy sell.

Oracle, on the other hand, was a startup with no other products. There was no internal case to be made - it was RDBMS or nothing.

Re: systems

John Cocke's early work on RISC went nowhere

Untrue. The 801 itself went from a research experiment to being used in 370 channel controllers and other embedded applications, and eventually the new core for the 3070's CPU. The 801 design was picked up as the basis for the ROMP, which was the CPU for the RT PC. The RT may not have been a big seller, but it was a commercial product and did enjoy some success.

Meanwhile, Hennessy and Patterson picked up Cocke's work at IBM Austin and created RIOS, which became POWER. That happened in parallel with the use of the 801 and ROMP. Cocke's work didn't languish in a closet somewhere at IBM; it was used more or less continuously following the first successful iteration of the 801.

IBM was then a place where good ideas went to die

IBM was then a place with a huge research organization. Much of the research did not become product, but a great deal of it did. Little things like relational databases (which IBM did, in fact, develop before Oracle), RISC, and virtualization. And a few non-computer things, such as electron microscopy.