On Thursday 17 February 2005 05:10, David Masover wrote:
> Adam wrote:
> | Hans Reiser <reiser <at> namesys.com> writes:
> |>Its on the legitimate wish list, if someone wants to code it, let me
> |> know.
> |
> | Hans, does this mean that you think that this type of functionality
>
> should be
>
> | implemented as a Reiser4 plugin and therefore in kernelspace?  Why
>
> wouldn't this
>
> | be better implemented in userspace via a daemon that is notified of file
> | modification via dnotify/inotify?
>
> Because dnotify/inotify don't scale.  I don't think they lock on event,
> either, whereas a plugin could guarentee that the hash was up-to-date
> (no race conditions).
>
> We've been over this before.  There's a reason reiser4 and its plugins
> are in kernel space, and not in something like Fuse.

And, surely, updating a hash value when 1 byte changes in a gigabyte file, 
would be much faster from kernel space where you can actually see the 
changeset?

-- 
Regards,
Christian Iversen

Reply via email to