1 Nov 2002 05:39
RE: BIND vs. non-movable resources in RFC3253, copy and cycles
Jason Crawford <nn683849 <at> smallcue.com>
2002-11-01 04:39:21 GMT
2002-11-01 04:39:21 GMT
Even in the absense of cycles, if you are doing a copy, and there are
diamond graphs,
a
/ \
b c
\ /
d
/||\
And d is actually a subtree, if the server does't reconstruct the diamond
or flag the duplication,
you're going to duplicate the sub tree below d. That might be expensive.
Do we have to require folks to reconstruct the graph? Probably not, but
we should
at least educate them about situations like this and encourage them to
recreate the
graph.
J.
------------------------------------------
Phone: 914-784-7569
.
The server can't avoid to track his inodes on a deep COPY to be
aware of cycles. So it also can easily find diamonds which will
exist for a reason I guess.
I think it's dangerous to destroy these diamonds by duplicating
resources.
Usecases which depend on the diamonds will break. So at least a
means to keep them should be provided.
Therefore IMHO disallowing duplication is the clean and
also spacesaving solution.
Cheers, Edgar
RSS Feed