Reading data from a file system is much easier then writing.  It's kind of like saying, I can extract the data from a database raw, but just because I can do that doesn't mean I can take into account all the undocumented nuances that I am ignoring to get the data in the first place, that the original system will consider corrupt if I miss even one bit in the proper place.

I can almost certainly extract data from most anything in a reasonable time frame (not encrypted) but it takes hundreds of times longer to figure out how to write that data in a manner that won't trigger a issue.

Steven

On Tue, 2003-10-21 at 15:05, Szakacsits Szabolcs wrote: 
On Tue, 21 Oct 2003, Jan Kratochvil wrote:
> On Tue, 21 Oct 2003 17:41:19 +0200, Szakacsits Szabolcs wrote:
> > 
> > Why not? NTFS has all the features that other Linux filesystems have and
> > even more. Why couldn't it be used as a standalone filesystem? 
> 
> As its data structures are undocumented. 

Lots of people spent lots of time many years ago to figure them out. There
are at least 4-5 different ntfs implementations based on it. Most of them
is read-only but if you know how to read, sure you can write as well. But
due to the concurrenct access, it's much more difficult to implement.

> And it IMO does not make sense to fork its documented derivative out
> of it (there was already one with unhappy end).

Sorry I don't get what you mean. The old NTFS driver? It didn't check the
NTFS version so when Microsoft improved it slightly (Win2K) and thus
updated its on-disk version number then the old driver tried to use it as
an NT4 NTFS. Trivial driver bug. Unfortunately nobody fixed it for a very
long time thus it ruined many people's filesystems. Thus because of the
Linux driver bug Microsoft became even more evil, NTFS completely
undocumented, later on the patent rumours added and etc.

The rewritten drivers and ntfsprogs check the NTFS version and exit if
it's unknown.

BTW, it would be interesting if one could check, try out what's Longhorn's
NTFS version numbers (i.e. if it changed or not). E.g. ntfsresize -i /device
would tell it.

> > Let me tell an example, the (Linux) NTFS driver supports transparent
> > compression.
> 
> OK, it is a feature missing in current GPLed (incl. specification)
> filesystems. It still makes more sense to me implementing the feature
> to existing GPL filesystem instead of implementing the same feature to
> NTFS with uncertain data structures. 

It's not uncertain, it's know for a while. Moreover NTFS isn't only a new
filesystem. It's also an important interoperability issue. Think about
FAT. Is it good there are all over open source FAT drivers? It even made
possible to quickly adopt to X-BOX's FATX. Soon FAT will go away
completely, only NTFS stays. 

> But we are talking about free software - do what you get paid for.

I don't understand this. There are many reasons why people write open
source software. Charity, bust ego, religion, fun, get paid, etc.
 
	Szaka