You're right, I didn't follow your argument completely but rather only a short subset slamming Unix. I get your gist, though I would have to see statistics to see whether making such required updates wouldn't reduce performance too extensively. I made the mistake of assuming you were a newbie who just didn't like the idea of disk caching period because the idea was problematic to your needs. I apologize for that error. I really didn't need the history lesson, at least the part about it that rubbed it in my face, but I'll live with it since I admit I was a bit condescending (I think you have returned that favor in spades though). Like you I have been in this business for quite a while, I have just concentrated on different areas. Finally, while I applaud attempt to improve Unix or any other OS, I would hesitate to imply that somehow something so drastic as what you suggest simply "stupidly" escaped the view of so many intelligent people who have contributed to Unix. It is I suppose possible they just all got caught up in a prevailing mindset, but I suspect it is more likely that there were *real* reasons at the time that made their now incorrect decision the right choice then. - Matt Richard Troy wrote: > > > If "rtroy" supposition was held up disk performance would suffer > > dramatically. > > This is the very old and very tired dogmatic argument which has been > disproven repeatedly throughout computing history. ...It could also be > that you have not understood - perhaps I have not articulated - the > argument correctly. > > > All OS-es use write caches, including Windows, though you > > have the option to turn it off. Not using caches leads to poor disk > > performance, especially on multi-user systems, disks with small buffers, > > and where you are regularly rewriting the same area. > > This is not a discussion about to use or not use disk caches. Of course > disk caches are wonderful things. The point is to not allow a condition to > occurr in which the STRUCTURE OF THE DISK CHANGES _without_ that change > having been written completely to disk prior to permitting user access > occurr. A properly managed cache knows the difference between things it > can write on convenience and things that should be written "now." > > > Also note that applications also use write caches, so even if your OS > > didn't cache and you powered off the system you would probably lose data > > with the application cache. If you really want data written immediately > > don't use "stdio" and do set O_SYNC when you "open()" the file. I've > > done so when I've had to and the performance difference is enormous. > > This paragraph clearly shows that you have not paid attention to my > statements. Either that, or you're unfamillar with the difference between > disk structural changes and the data contained therein. A previous poster > had it right when he wrote something like "disk stucture = meta-data". > BINGO! We are _not_ talking about user-data - at least, I'm not. > > > If Unix is stupid because of this, all modern computers are stupid. This > > is standard computing practice, Bill Gates included. > > Well... It's true that Unix, up until relatively recently, has had this > rather stupid flaw. They band-aided the thing with fsck - an embarassment > in my view - because it was _required_ with their strategy. It sucks. The > argument was that, "these are research systems." And, indeed, most were. > How do I know this? I was there. It's a very sad truth that computer > science has suffered greatly because of a mis-guided perception that > anything older than about 5 years is ancient history and not worth > learning. There have been no less than three complete generational > changes - epochs through which similar or identical lessons had to be > re-learned because of this short-sightedness. Note, we still use the wheel > even though it's more than 2000 years old, though we have improved upon > it. Too bad we haven't done so well in the short history of computing. > -shrug- > > Anyway, as for Windows, it always did suck. I remember when it first hit > the streets. I took a serious look because at the time I was writing my > own operating system to control ships, refineries and pipelines. It was > immediately obvious to me that it was antiquated before it even started. > One of the most obvious flaws was that it didn't have a proper inturrupt > system - and that remained true until about Windows 95 - which came out in > '97 if I recall correctly. It was only when Dave Cutler, inventor of > RSX-11, and subsequently the primary architect of VAX/VMS, went over to > the dark side when Ken Olson snubbed him, demanding he close DecWest (in > Seattle)... Ken made the big mistake of letting him and his code go. > ...That code line became Windows NT... > > As for "all modern computing" - well, all modern computing does not > include some really wonderful technologies which were created. These have > been largely lost due to two primary factors: The failure of many big > computing companies to correctly handle the microcomputer market, and > Gate's illegally won monopoly. Sure, there are other reasons, but it's a > huge mistake to equate modern computing with the most intelligent > computing. > > I'm guessing that all of this is unfamillar to you. I lament that. That > old quote is true, I think, the one that says that those who for get the > past - or who never learned it - are condemned to repeat it... > > Regards, > Richard > > _______________________________________________ > Redhat-devel-list mailing list > Redhat-devel-list@redhat.com > https://listman.redhat.com/mailman/listinfo/redhat-devel-list -- --------------------------------------------------------------------- Matt Fahrner 2 South Park St. Manager of Networking Willis House Burlington Coat Factory Warehouse Lebanon, N.H. 03766 TEL: (603) 448-4100 xt 5150 USA FAX: (603) 443-6190 Matt.Fahrner@COAT.COM ---------------------------------------------------------------------
