David.Collier-Brown at canada.sun.com
Wed Sep 20 16:27:01 GMT 2000
On Wed, 20 Sep 2000, David Collier-Brown wrote:
> > Should I assume that not supporting change notify
> > speeds this up? If so, can it be turned off
> > without turning off any other usefull features?
Jean Francois Micouleau wrote
> Dave, IMO the correct solution is to have an underlying OS able to send
> change notify to samba and in turn samba propagate c.n. to the client.
I've looked at the mechanisms in Samba and, IMHO
1) it can be improved somewhat
2) it **might** be overkill
3) the system could do it better.
Looking at these in reverse order...
3) In Unix v6 (and Solaris), it's not
obvious where one would add the appropriate
dirty bit: gut feel says in a extended header
for a directory in cache... However, this is based
on only 30 second's analysis (;-))
2) The semantics of "change notify" on a directory
**may** be only those things which the directory's
last-modified time represents. However, this is
a speculation based on looking at the operations
accessible via Windows explorer. Similarly, change
notify on a file, if defined, might map to the
Anyone who groks NT: does this make sense? If not,
what are the operations that aren't subsumed, and could
customers live without them?
1) We should cache the directory mod time and check
it first, as a necessary-but-not-sufficient test.
This could allow us to avoid doing the whole readir/
stat/checksum traversal in cases where the directory
alone was modified.
David Collier-Brown, | Always do right. This will gratify some people
185 Ellerslie Ave., | and astonish the rest. -- Mark Twain
Willowdale, Ontario | //www.oreilly.com/catalog/samba/author.html
Work: (905) 415-2849 Home: (416) 223-8968 Email: davecb at canada.sun.com
More information about the samba-technical