virtual memory 4510

Anne & Lynn Wheeler

It seems unlikely that working-set implementations were anywhere nearly as mature as LRU was at that time, doesn't it? I doubt that anyone would suggest that working-set approaches are anything but butt-ugly by comparison to something as simple and elegant as global LRU, and chock-full of heuristic tweaks - but that's not the point: the real question is whether they work better for heterogeneous and occasionally partially pathological workloads over a wide range of physical memory sizes.

... and as referenced, it was a decade later

Just how much is there to 'work' on with global LRU, anyway? A few of your statements seem to have suggested significant extensions to address some of the problems that working-set approaches are designed to address - if so, it would be informative to know more about them.

.... although there was

Dear me, that sounds so much like Russian academia in the bad old days was allegedly like that it's difficult to take seriously. Not, mind you, that I discount the existence of compebreastiveness and blatant politics in American academia - I just doubt that it would be very effective in suppressing useful work in other than occasional specific instances, given the number of available venues through which it can appear and their own compebreastiveness in presenting material of significance and interest.

That's getting repebreastive, I'm afraid. As I stated earlier, I'm interested in head-to-head studies using repeatable traces rather than anecdotes with far less stringent controls on several significant variables.

- bill