> >> > Only once fully formated (which > >> >takes 1 hour for 1x media), DVD-RW permits for random write, but with > >> >32KB granularity. Latter means that you either have to have special > >> >kernel driver which would arrange for 32KB granularity or modify file > >> >system code to do same thing. Now note that no work (at least no visible > >> >outcome so far) has been done to accomplish either of these two > >> >alternatives under Linux (or any other Unix implementation), [presumaly] > >> >because it's damn hard. > >> > >> So DVD+RW drives do things in firmware what you need to do in the kernel > >> if you like to have 2k granularity with DVD-RW. But this is nothing a new > >> firmware could not support. > > >Specifications are very explicit about I/O granularity in DVD-RW > >Restricted Overwrite mode. You seem to be willing to bend standard to > >suit you, it's not fair play. Secondly it doesn't really matter what one > >*could* *possibly* do [to stretch something to something else], does it? > > Well, it seems that you make unproven assumptions on the quality of DVD+, so
We *both* have only one way to prove our statements: take vendor's word and try to refute it by practical experience. If vendor's word holds the scrutiny, then it's considered to be proven. Now, when I say that DVD+RW implementations provide for 2K granularity and true random access after minimal formatting procedure, I'm not just repeating something I've read, I speak from practical experience, both my personal and of several users. > I believe that it is fair to tell people that in theory it would be possible > to > enhance DVD- firmware Yes, it's obvious that it's *theoretically* possible to implement 2KB write granularity in firmware [once again, which would violate DVD-Forum specification]. > and hardware ^^^^^^^^^^^^? > so partially formatted media may be adddressed ramdomly. Well, once you call my statements unproven, I think I can demand proof for this one of yours. Prove it to us, that *minimally* formatted DVD-RW media can be written to in truly arbitrary order. Or at least explain why there is no implementation, which would actually provide for this. > >> Note that the kernel also first reads a 512 byte sector from a hard disk > >> if you like to write only 64 bytes. > > >Yes. But as already implied, if you want to extend this to 32KB to > >accomodate DVD-RW Restricted Overwrite, you have to modify kernel file > >system driver. The question was "why is there random access possibility > >for DVD+RW, but not for DVD-RW?" > > If you believe that the kernel filesystem driver needs to be modified, then > you are taling about a broken OS. Oh! You must be implying that I'm referring to Linux. Well, I'm not. I'm referring to a typical OS and we can speak about Solaris if you wish. Have you ever tried to build a file system under Solaris with logical block larger than 8KB? Try! And before you start arguing that it's user-land mkfs_* which set this limit. Yes, user-land tools set the limit, *but* for damn good reason. Because the limit is hardcoded into filesystem kernel drivers as well [I've examined ufs and udfs code]. This is [*one* of the reasons] why I wrote "you have to modify file system code." > On a typical OS, the filesystem code talks to the block abstraction layer. > > This layer has been the buffer case on historic implementation and is the > semgment driver layer + VM Cache in modern implementations. > What you have to do is to change this layer to know that there may be > 32 KB Blocks..... The question was "why is there random access possibility for DVD+RW, but not for DVD-RW?" The question also was in certain context and can as well be rephrased as "why is there possibilty for random access of DVD+RW under Linux now, but not for DVD-RW?" Your answer essentially is "there is random access possibility for DVD-RW as well, *if* one a) modifies unit firmware, b) modifies unit hardware(!), c) modifies OS VM cache." Pure children's game... A.
