Regarding TransRelational — the patent and other public descriptions generally refer to a memory-centric, query-heavy design. Actual development adapted these ideas for a more disk-based, update-friendly approach, in a columnar architecture. (Pascal can fulminate against my use of that word all he likes, but it’s well-founded, given the caveat that “columnar” can mean many things by the time the system is finally implemented.) The memory-centric version is pointer-crazy — one of the big virtues of putting things in memory is that you can use pointers freely — so it can’t be translated to disk without heavy adaption. I don’t know exactly what changes they made to the publicized ideas. 

It’s a pity that the CEO and the investors at Required couldn’t get along. I don’t think the problem was the technical bait-and-switch (building something different from what they said they would when they raised the money); that happens all the time. Rather, it’s the classic issue of a founder refusing, rightly or wrongly, to be replaced by “professional” management. I do not have a strong opinion as to who was “more” at fault. I do have the strong opinion that the technology is/was NOT so radical as to easily survive a three-year shutdown and ongoing corporate dysfunction and yet wind up with an “orders of magnitude” breakthrough coming to market. 

And I’ll stop at that, because while I don’t use the full journalists’ playbook, I do have an issue here with protection of sources. 

Comment by Curt Monash — November 1, 2005 @ 7:11 pm 
http://www.dbms2.com/2005/10/29/oh-dear-chris-date-is-displeased-with-me/