In message <199906101017.daa27...@mag.ucsd.edu> Bill Huey writes: : It's a good thing because assumptions about memory useage within the kernel, : portability abstractions, internal buffer queue overhead, etc..., need : to reexamined to see if they are still relevant.
While it is true that one needs to occasionally reexamine the assumptions that one has built one's system upon in the light of new information, the mere fact that one does so doesn't mean that how things have been done in the past is automatically wrong. Also, there have been years of research in these areas so you'll know what to expect as the CPU speed gets faster, memory gets cheaper, etc. It isn't like the people that have done the years of OS reasearch are stupid and short-sighted. Yet, this is what the message that I've gotten from the Linux folks in the area of kernel development. Maybe it isn't intended, but it sure is offensive. : This is always good, assuming that this is done properly with peer review : and that folks listen to it. Linux isn't peer reviewed in the traditional sense of this meaning, so your whole argument fails because of that. I'd agree if it was entensively peer reviewed, it might be a good thing. It isn't. Linux gets to where it is going by doing things many times quickly, but not necessarily correctly. While there is some value in this approach, it can be quite wasteful. I gave up maintaining my port of Linux to my rPC44 because the internal internal interfaces kept changing at a rate that makes the recent -current newbus integration look and feel stable. Warner To Unsubscribe: send mail to majord...@freebsd.org with "unsubscribe freebsd-hackers" in the body of the message