"Expert" witnesses for US courtrooms is a special kind of a parallel voodoo-science world. Especially when it comes to arson.
Prosecutors like to pick the same people to testify as "experts" and their top qualification is that they have testified before as "experts". I imagine many have optimized putting up an act and throwing around fancy terms to make it seems really precise and scientific. Their future employment depends on that.
In civil cases, where both parties often have a significant budget for hiring experts, the experts often actually are top people from their fields but who are retired.
I watched a patent trial, and an important part of one of the claims was that certain data from a hard disk was cached in a RAM cache.
The side that wanted to see the patent invalidated brought up some earlier patent that involved caching some data on a hard disk, and claimed it was prior art that should have been considered. Their expert said that "RAM" stands for "random access memory", and a disk is random access and it is memory, so that any programmer or engineer who read the patent that called for using a RAM cache would understand that caching that data on a hard disk would count.
That, of course, was one of the most ridiculous things I'd ever heard. However, the expert certainly was a real expert. He was a professor emeritus and former head of the computer science department of a top CS school. In fact, he had been the one who started the CS program there and brought it to its position as a top program. He was a fellow of the IEEE, and his published papers back from when he was an active researcher were some of the seminal papers from the beginning of the modern semiconductor and VLSI era.
So how the hell did this guy think that "RAM cache" is commonly understood to include a cache stored on a hard disk?
I asked the lawyers for the side that was defending the patent, and their expert (another retired professor emeritus, IEEE fellow, yadda yadda...) about that. They explained to me that experts in a trial like this will get paid $50-100k or more to testify in this kind of trial. They basically have to just spend a few hours reading a patent, and then come up with an argument for the side that is paying them that will sound good to a jury and that is not provably a lie. Then they have to come and spend maybe a week at the trial location, being there to help the lawyers out and to spend maybe an hour on the stand. So basically $100k for a couple of weeks of not too strenuous work, two plane trips, and a week in a hotel.
They don't have to worry that colleagues will find out about the dumb things they said at trial and hold it against them or that it might tarnish their legacy or reputation, because everyone in the field knows about these expert witness gigs. When they hear that Professor Soandso said something really stupid about RAM cache at a trial, they just think, "Oh...I guess Professor Soandso wanted a new RV" or "Professor Soandso wanted a new boat". (In the case of the expert for the patent owner, it was to be one of the earliest to own a Tesla).
I think you're being too uncharitable to the expert. It sounds like his point was that caching on disk is conceptually similar to caching in RAM, and that therefore, for purposes of prior art, you should probably list inventions that cached in this (similar) way. IOW, an engineer reviewing the literature should have included inventions that cached on disk if they were otherwise similar.
Remember, prior art is intended to demonstrate the extent to which your idea is new. It makes a big difference whether the inventor thought of caching at all vs whether they just used an existing caching scheme and swapped out the implementation for a faster one.
Posters on HN make a big deal about obvious patents being granted because someone took an existing invention and added "...on a computer/the internet". Well, it's a similar problem when someone takes an existing mechanism and says "... but on a faster storage medium instead".
> So how the hell did this guy think that "RAM cache" is commonly understood to include a cache stored on a hard disk?
Maybe because he really did? To paraphrase you: How the hell did we arrive in 2016 where almost everyone believes "map/reduce" implies parallelism or where the inventor of the term "object orientation" permanently has to set the record straight about what he really meant and how far our understanding today diverges from that?
Prosecutors like to pick the same people to testify as "experts" and their top qualification is that they have testified before as "experts". I imagine many have optimized putting up an act and throwing around fancy terms to make it seems really precise and scientific. Their future employment depends on that.