Re: Hashing for DISTINCT or GROUP BY in SQL

On 15/10/2010 2:54 PM, Anne & Lynn Wheeler wrote:
>
> paul c<anonymous_at_not-for-mail.invalid> writes:
>> Later I met an Amdahl salesman who said, "give me more of this
>> relational stuff, I can sell all the 'boxes' it needs". When a boss
>> of mine plunked for an Amdahl cpu, the local IBM field manager invited
>> him over for coffee and 'career counselling'. That was in the
>> mid-1980's. By the early 1990's even IBM could see which way the wind
>> was blowing and introduced the RS6000, basically a Unix machine, an
>> attempt to hedge their bets, having already blown their early PC
>> lead. In other words, just like other people do now, they knew
>> something big was happening, they just didn't know what.
>
> lots more complicated than that.
>
> I knew many of the people at amdahl ... including the guy doing the
> amdahl dbms HURON.
> ...

A Dane named Knudsen, brilliant with radical ideas about how to structure development but secretive (he had spent time at San Jose on a fellowship, personally I think much of his personality was formed by events just after WWII when the effects of the occupation were remembered and still being felt). He spent much of his time out of town and the way he kept control was to make the interfaces between components inviolate. It was amusing to watch him make fun of the EDS people when they bought in. Everybody laughed except them, not because they were insulted, but because they usually didn't get the joke. Actually, it was a very nice man named Bill O'Connell who initially funded it, he owned the AT&T business (when AT&T was at 'war' with IBM, so they bought all their mainframes from Amdahl, the NSA was also a big customer even though Fujitsu was majority shareholder) so he had a lot of discretionary budget. He called himself a salesman but for a salesman, he had his name on a lot of patents. It was a great pleasure to go out to dinner with Bill, nothing but the best and lots of it. Amdahl planned his new cpu's on O'Connell's whiteboard (or was it a blackboard then?) before they both left IBM.

Huron's original target was Cobol programmers but that morphed after a while to a kind of legacy support tool because the original plan included interfaces to pretty much every other dbms, albeit solely for migration purposes. It might have been the first large 'RAD' product, extremely simple to use and teach, with great tpc numbers on the mainframe and with unix and windows versions whose performance surprised people but it was definitely a 'big-ticket' product. It had explicit 'forall' but on the surface, most people saw a row-interface but that didn't stop Codd from being interested. Knudsen didn't want to give up his title. I often wonder what might have happened otherwise. The logical access method was called TAM (for 'table access method') and the physical was called NAM ('native access method'). When the product went public, the Yank shareholders thought the latter label would remind customers of an unhappy war so the component name was changed to 'tds' (table data store). That was typical of the minimal understanding by management of how radical Huron was. Among the Huron developers only one had anything approaching a deep understanding of Codd's ideas but he was prevented from working on any of the logical aspects by agreement with IBM where he had previously worked on the sql/ds optimizer.

Huron programs were called 'rules' and were tiny, limited to less than 4K bytes each. They had a canonical form and there was no such thing as 'source code' after the first compile, 'source' would be generated from object. The structure of a rule gave people the impression of a truth table, there being no 'if' statements per se, but that was a bit of an illusion. It was amusing to hear prospects object saying, "but my Cobol object deck is 20K, how can I re-write that as a Huron 'Rule'?".

There was a big Texan you'll remember named Tom Simpson who was previously known for creating HASP at IBM and who used to travel up to Toronto to give advice. There was often talk of taking parts of Huron in other directions and one idea was to use some of it as a system programming tool. When Simpson asked, "well, just how would you issue an I/O?", he was taken aback with the answer "oh, that's just a table insert". (ie., persistence wasn't considered a logical requirement unlike in other dbms'.)

The first couple of years after introduction were quite promising but then new management came in and mangled the original ideas, not even being aware what they were stepping on, going so far as to jump on to the OO bandwagon without knowing exactly what that was nor what it would it lead to. When they moved QC to Dallas, all hell broke out at customer sityes. Amdahl and EDS management also went too far, overselling and trying for too many customers too fast. Many, many marketing and support people were hired, most of them knew nothing of db, very little of typical large business apps and had never programmed and promised customers the moon when they themselves were incapable of delivering any of what they promised, ie., management hired people exactly like themselves. The avalanche of use of pc's by business in the mid 1990's put 'paid' on the original mis-steps. Like lots of companies they forgot their market. Received on Sat Oct 16 2010 - 02:22:07 CEST