willy@thepuffingroup.com (willy@thepuffingroup.com) writes:
> On Wed, Apr 05, 2000 at 11:41:18AM +0200, Pavel Machek wrote:
> > Yes. I'd like UFS going userspace. Anyone is going to submit nice
> > midnight module? FAT module for midnight would be also nice, as it
> > would allow accessing floppies from userspace; and broken floppy would
> > no longer crash your machine. (NFS is used too commonly, and _is_
> > performance critical).
>
> UFS should stay in the kernel; it is a _sane_ filesystem, designed for
> Unix, and it's at least theoretically possible people will want to use
> it as their root FS once write support is fixed.
Did I miss something? "Should" stay in the kernel? It doesn't make much sense for it *not* to be. This whole idea of user space filesystems is evil, anyway - the only reason you'd want to do such a thing is to allow users to implement something they normally don't have permission to do in the first place (and for good reason, I think), or to implement something to test or something that most people won't need (or want). Things like CFS over loop aren't bad ideas, but they suffer performance-wise because they *are* in user space. The only reason they're not in the kernel is because there's not a whole lot of demand for them (but more likely because Linus hasn't, and probably won't, sign off on them). Personally, I think it would be a cool idea to implement crypto in the VFS layer - then you wouldn't have to worry about hacking drivers for each little file system. But, since you've got loadable modules, you can pretty much do what you want.
But just because you *can* do a thing doesn't mean you *should*.
I also don't understand the contention that implementing a FAT module for midnight would allow you to "access floppies" from user space - you can do that now, anyway, and always have. Open /dev/fd0, do your thing, and go. The main advantage to raw devices is (1) to support legacy code, and (2) to allow sync reads and writes to the device without having to do O_SYNC hackery and such. And what's wrong with NFS, aside from the fact that it's rather slow (well, what isn't over a 10MB wire?), and doing attribute caching is a hassle for the user?
Or have I just not had enough coffee this morning? ;)
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Apr 07 2000 - 21:00:18 EST