(Illustration by Gaich Muramatsu)
On Wed, Jan 14, 2004 at 01:03:52PM -0500, Greg Troxel wrote: > I downloaded (on a well-connected machine) a file to a dir in coda, > named cfs.pdf. > I looked at it on a poorly-connected client (long wait, worked ok). > > On the fast client, I did 'mkdir FS; mv cfs.pdf FS'. > The slow client got an invalidation callback. > > I then did 'cd FS; md5 *'. The client fetched cfs.pdf anew. > It would be really nice if it could have kept the bits and associated > them with the new file, either by SHA1 hash or an explicit tracking of > the move. > > Actually, I would have expected this to work, since there is no reason > for the fid to have changed. Or does the move change the VV or > storeid? Correct, I wonder why it refetched the file. I can see it might refetch the attributes because the move operation possibly also invalidated those (mtime/ctime), but as the object itself hasn't changed it should not have to refetch the data. JanReceived on 2004-01-14 17:28:49