CompuServe Thread

#2090 HARDDISK CONTROLLER

16 messages in this thread
#96946From: Ben BlishDec 6, 1987 12:22 PM
I have a sneaking suspicion that there is a "ceiling" around 60k or so, because of DOS. Currently, anyway. As for the new file system, We're all still waiting, and we (SoftCircuits) are still tweaking our own. As is Perry, If I have his insinuations right… <grin>. –Ben–
#96968From: John DraperDec 6, 1987 2:10 PM
Ben, I keep hearing about faster file systems from you, but wonder if it's the way to go. it would seem to me that we need one file system, not 2 or 3. Without standardization in this area, how will anyone write anything that uses it? Regards, Larry.
#97273From: Dave HaynieDec 8, 1987 12:47 AM
That's the whole idea of handlers! Unless you're writing something like DiskSalv, where you really need to diddle with the structure, a file handler's particulars should be local to that handler, and not at all obvious to the outside world. In fact, if you can tell the difference, someone's not doing their job in building their handler. I agree that for floppies you need a standard file system to transfer data between machines, but why on hard disks? If your file system adds speed or something, that's the only difference you'd notice. Now if you start adding special packets or something, then you get into specific handler support, but those kind of things are best left to the supplier of the filesystem, and anyone who wants to hack with them, rather than a thing you'd like to support in a normal product. -Dave
#97284From: John DraperDec 8, 1987 1:39 AM
Dave, I think you answered your own question when you said "Unless you're writing something like DiskSalv". Since the handler is what determines the contents of the blocks on disk, a different layout will break programs designed to work with another layout. Sector editors, file recovery programs, disk mappers, etc. will all require a knowledge of the mechanics of storing the file. A good example is Sectorama, which only knows about the current file system. It can still be used, but only in a limited and not overly friendly way. Regards, Larry.
#97351From: Ben BlishDec 8, 1987 4:46 PM
Larry; Would you say that you support the idea of not releasing a product that can triple (or better, in Perry's case) the HD throughput of an Amiga system, on the basis that SectorRama and it's brethren would no longer function? –Ben–
#97368From: John DraperDec 8, 1987 6:31 PM
Of course not Ben, i do, however, fully support having only one such product, and having it supplied with the machine. As with anything else, the one that comes with the machine is the one that the majority will use, but with just enough users of 'some other system' to throw significant confusion into it. Regards, Larry
#97683From: Ben BlishDec 10, 1987 4:43 PM
I don't see why. There are a bunch of copy commands running around; And no one seems to be confused. They all copy, they all work, and they're all more or less compatible. Some are higher performance, some are missing features, but they're all acceptable. Same thing. –Ben–
#97698From: John DraperDec 10, 1987 5:43 PM
Ben, it isn't the same thing at all. A copy program interacts with the OS on a level that virtually guarantees compatibility. This will not be the case with programs that, by necessity, need to go to the device level and know about the file/disk format. Regards, Larry.
#97771From: Ben BlishDec 11, 1987 12:19 AM
Only repair programs need to do that — not "normal" application programs. A file system interacts with this OS on a level that is just as standardized as the DOS level; DOS packett level. If you describe a proceedure for "cleaning up" a disk using the AmigaDOS copy commands interactive mode delete selection, then a unix-style replacement is not only incompatible, but sadly deficient in comparison. If you should try and clean up a disk partition with a tool not meant to handle it, then you suffer… If the tool, on the other hand, uses the file system in the prescribed manner, then it should work. I suggest that all of us should be writing our stuff at that level. Otherwise things don't work when the OS gets revved. As we have seen. –Ben–
#97794From: John DraperDec 11, 1987 1:47 AM
Ben, Repair, diagnostic, and investigative programs are exactly what I am talikng about. You are talking about something at a higher level. Though higher level programs can be used, you must admit that there really is no substitute for a low level program that allows the grey computer upstairs to make decisions. Regards, Larry.
#97822From: Ben BlishDec 11, 1987 8:47 AM
I admit it – but I don't see it as a sufficient reason to forego a significant performance advantadge in my work… The few times that things have gone sour on me, usually due to software in the pre-release stage (read, my fault) things have been so badly bit in the chops that only many hours with those utilities could fix even a fraction of the problems. The average user not only is in a position where they don't have such utilities, they are unable to use them if they had them. Whereas, if I were to release some software that gave you, say, 250k throughput over the current or then-current speed you were getting, I bet you'd dig in and MAKE the utility you wanted. Or the whoever it is who makes those fix it type of programs. Not me, that's for sure. –Ben–
#97338From: Ben BlishDec 8, 1987 4:18 PM
Larry — it's (ours, anyway) totally transparent to the software that uses it, unles said software uses direct-to-sector techniques like some games (cp'd) and some utilities. Any software that is regular application software, follows reg'lar dos rule, won't know it's running from our FFS. Only you will know, because it's faster. There is no effort required from anyone to support it, _except_ those who write disk-doctor type utilities… because they _care_ about the disk's actual sector-to-sector structure. –Ben–
#97366From: John DraperDec 8, 1987 6:27 PM
My point exactly Ben… let's see… was this Benhandler or Perryhandler format?
#97682From: Ben BlishDec 10, 1987 4:42 PM
Or CBMhandler format… That's probably something that will add value for those who write such things; It's not a good reason to skip the change. In any case, because of the way it's implemented, you _do_ have a choice. You can keep the old file format/handler, and run at your usual speed with all your disk tools intact; Or, you can go with the new format, give up your disk tools, and run at 3x your old speed – or better. That's going to be an easy choice for most people, either way. –Ben–
#97413From: Glenn NielsenDec 8, 1987 9:16 PM
Ben, I also have reservations about having multiple FFS for HD's on the Amy. Expecially if there isn't a significant difference in performance between the FFS's. As Larry said, it will add confusion on using disk fix type software. Of course, if each FFS came with it's own disk fix utilities I would be appeased somewhat. BTW, what kind of performance are you getting with your FFS? Do you have some DiskPerf results you could share with us? Glenn
#97685From: Ben BlishDec 10, 1987 4:47 PM
Glen, I 100% agree. An FFS implementation has two things to offer us, as users. First, higher performance, and second compatability. I think that you should choose the system that works best for you, in both ways. As for diskperf figures, no, not yet. Still tweaking, and that makes a lot of difference. The Amiga is _very_ sensitive to little changes in format and so on; We'll release figures at the same time that we release the Beta stuff. We're kind of looking at Perry's new system and CBM for some goals — we're not sure just how good "good" is, if you follow me. –Ben–