Anders Norrbring wrote:
What are your NSHO ( ;-) ) about JFS?
Greg Freemyer wrote:
Since then I've tracked it (and it's bugs) closely, but I've never used JFS.
IBM's JFS has the same problems as ReiserFS, it was _not_ a traditional UNIX design. It is a common _misconception_ that JFS was ported from AIX. JFS was actually ported from OS/2. The reasons for this are more legal than anything. Long story short, IBM was prevented in a Non-Compete Clause in the Project Monterey Contract (with SCO) from porting any work from Monterey/AIX to Linux/IA-64 or any 64-bit enhancements to Linux, such as AIX's 64-bit filesystem. This was in 1999, _before_ IBM decided to pull out of Project Monterey in 2000, and basically undercut Caldera 20 days after purchasing SCO (and the rest of that is for common debate). So, instead, IBM ported JFS from its original OS/2 foundation. This mean it lacked many of the inode features and compatibility that the AIX version had -- for quota, NFS, XATTR, etc... support. As such, JFS also has similar compatibility issues on Linux as did ReiserFS. In fact, IIRC, it's largely SGI's donations to the Linux VFS (virtual filesystem) layer in early 2.5.x developments that JFS and ReiserFS now have _some_ quota, NFS, XATTR, etc... compatibility (long story). XFS was ported from Irix _directly_ to Linux with *0* removed. In 2000, XFS was already a filesystem with a 7+ year, proven and _unchanged_ structural design. That's why it is very much trusted and proven in Linux -- with caveats (e.g., the 2.4 kernel lacked a lot of capability required for XFS features -- hence the SGI donations to 2.5.3+). ReiserFS was not a traditional UNIX filesystem design -- let alone its structure changes too much. While the kernel implementation and on-line check might accommodate this, I've just had too many issues with the off-line tools being "out-of-sync" with the on-line organization. And JFS, from OS/2, was not a traditional UNIX filesystem design either. I haven't kept up with it post-Monterey fall-out to see if IBM has changed its stance. As much as people argue back and forth about SCO's claims on IP in Linux (which I won't do here), the Non-Compete Clause of Project Monterey combined with IBM withholding all Monterey/IA-64 source code, is still the "main meat" of SCO's lawsuit (and not the "IP smokescreen" that was added in May when IBM didn't settle, after the original March filing). So it would not surprise me if IBM hasn't ported some of the AIX code, such as from JFS, to Linux because some of the rulings on Monterey-driven discoveries by judges thus far (again, _ignore_ the Linux IP stuff). And then there's also the fact that IBM still considers Linux to be the "younger sister product" to AIX, where they still have a lot of sales and profit -- one of the reasons they undercut Caldera once they purchased SCO (who was going to introduce the same split UNIX-Linux strategy -- and the Non-Compete Clause in Monterey being the whole reason Caldera thought they were "safe"). -- Bryan J. Smith Professional, technical annoyance mailto:b.j.smith@ieee.org http://thebs413.blogspot.com ----------------------------------------------------------- Americans don't get upset because citizens in some foreign nations can burn the American flag -- Americans get upset because citizens in those same nations can't burn their own