Re: OpenBSD and FS-Support (compiling kernel without support..)

[Chuck Yerkes <chuck+obsd@2003.snew.com>]

Quoting Thorsten Glaser (tg-v2003@netcologne.de):
> Dixitur illum chuck+obsd@2003.snew.com scribere...
> 
> >I've never been really clear why non-native file system support
> >isn't done via a kernel module.
> 
> Because this is not the (Open)BSD philosophy.
> We don't need to modularize the kernel. It's good as it is.

"change is bad" - I like it.

Nice view.  I've got 4.2BSD on QIC60s somewhere if you want
to stick with that.

Quickly:
kernel modules let me:
- not load things I don't need (union, ext2fs, several dozen other
  things) - to let me run a smaller kernel
- load modules which might have GPL or other non-BSD licensed code
  (which a reiserfs modules would likely have).
- load and unload modules while debugging, rather than 
  code, compile a kernel, boot, hope it works, debug.
  RATHER: modload module, try new calls, hopefully modunload.

Sometimes it eats it and you reboot with either path.

/usr/ports/kernelmods/  would be a fine way to carry non-BSD
licensed modules and maintain the integrity of the goals.

> -- 
> [...] if maybe ext3fs wasn't a better pick, or jfs, or maybe reiserfs, oh but
> what about xfs, and if only i had waited until reiser4 was ready... in the be-
> ginning, there was ffs, and in the middle, there was ffs, and at the end, there
> was still ffs, and the sys admins knew it was good. :)	-- Ted Unangst ?ber *fs

Right, we can list the shortcomings of FFS.  Esp FFS1.  Snapshots
in FFS2 let me do really cool things, now we just need fast directories.

Ever have a machine with 1TB mounted go down hard?  With XFS (and reiser
and VxFS, and others), there is no fsck.  There is no need for fsck.

find / >/dev/null
   takes  < 1second on reiserfs and around 10 minutes on FFS and HFS+.
   (similar number of files).