On Wed, Feb 26, 2003 at 05:18:42PM +0200, Ira Abramov wrote: > Quoting Muli Ben-Yehuda, from the post of Wed, 26 Feb: [snip] > > You forgot to mention the most important thing, the reason why: AFAIK, > > the NTFS "on disk" format is undocumented and what is known about it > > has been revesed engineered. > > in essence it was supposed to be quite similar to the OS/2's HPFS, but I > guess it's different now.
Actually, the details are undocumented, so you really have almost nothing if you want to implement it, but the high-level ideas are documented in a book named "Inside NTFS", which is quite interesting, and which made me (when I read it, many years ago) appreciate NTFS quite a lot. My current opinion on it, similar to my opinion on many things MS (and others), is that the design was quite good, only the implementation sucks. In any case, it's not "similar to HPFS": it's much better, and has some very good ideas which were new at the time, such as no fixed size structures - that is, you do not (theoretically, I do not know what happens in practice) need a ntfsresize (similar to ext2resize). One idea it does share with HPFS is that it has indexes. HPFS has one index, on the file names, so that a query on a specific file inside a huge directory should be much faster than other FSes, and NTFS has a potential to index everything (such as search for files in specific size or date ranges), with implementation for name (at NT4, don't know today). Why do I say the implementation sucks? Simply because during the years I had too many times (maybe 10-20 by now) NTFSes which were currupted to the point of hanging autochk (which runs at boot, and therefore stopped NT from loading), and at least in some of the cases I managed to rescue most of the files with the unsupported, reverse-engineered driver of Linux. > > If I must install such a dual boot machine, I give NT its' NTFS for the > system and leave a big partition in fat32 for all to share. but I try > not to get to that situation in the first place :) > > > -- > The man with the golden .sig > Ira Abramov > > http://ira.abramov.org/email/ This post is encrypted twice with ROT-13. > Documenting or attempting to crack this encryption is illegal. > Didi ================================================================= To unsubscribe, send mail to [EMAIL PROTECTED] with the word "unsubscribe" in the message body, e.g., run the command echo unsubscribe | mail [EMAIL PROTECTED]
