[CrackMonkey] Uh...

[Mike Goldman]

Monkey Master wrote:

>         Journaling filesystems, however, are rather slow.

Checkpointing is slow, but by and large a journaled filesystem should be
faster than a traditional fs.

>         Every time anyone has mentioned softupdates on the linux
> kernel list, people have said "So what?  ext3 will have journaling".
> This is a stupid move, and shows ignorance of what softupdates really
> does.

Well, ext3 won't be in 2.4 most likely.  ReiserFS will probably be,
including journaling. Since this is a whole lot different from traditional
fs design in ways much more significant than the addition of journaling, it
would be hard to isolate just that aspect for comparison.

It seems to me that softupdates is a form of "journaling on demand."  If
you think that journaling slows things down in the general case, and don't
want to enable it until/unless you need it, then this probably seems like a
superior solution.  But if journaling is only slow in the checkpointing
phase (which will be necessary to reconcile the softupdate, too), then it
may not seem an improvement at all.

I may be stupid, but I don't think softupdates are preferable to a
continuously journaling fs.