[<<]Message[>>] [<<]Author[>>] [<<]Subject [<<]Thread
Number : 276 Date : 2001-06-22 Author : Kan Yabumoto Subject : Re: XXCOPY New Beta Versin (v.2.56.5) Size(KB) : 4
Dear Chaz: I'm glad that the new /FF feature is working as I was hoping. But, of course, since the granularity problem as we know it should be always within 2.0 second (exclusive of exact 2.0 sec). That was the reason why we chose the default value of /FF to be +/- 2 seconds. Naturally, when we are adding a fuzzy factor, we should keep the range as tightly as possibly we can. So, technically, for a given set of file-system combination, the tightest possible value (the optimum setting) should be either /FF+2 or /FF-2. XXCOPY src/dst optimum parameter ----------------------------------------------- FAT-src --> NTFS-dst /BI none NTFS-src --> FAT-dst /BI /FF-2 Unfortunately, it is not as simple as it seems like this. The granularity truncation takes place when a file is copied from a finer-time volume to coarser-time volume (e.g., NTFS --> FAT or UNIX --> FAT). And the trouble is that some of the conversions are not necessarily to discard the fractional value (as we programmers consider as most logical and easy), Microsoft's built-in conversion routine actually adds fractional value to make it "newer" than the original timestamp. That was why our earlier efforts in this area were not always successful. And, to be honest, we have not figured out exactly how some of the time discrepancies are introduced in the first place. --------------------------------------------------------------- Of course, we are in a danger zone here. The first time we saw Robocopy's built-in 2-second relaxed time comparison, we did not like it. We wanted to be more exact --- our earlier efforts in the /FT and even the /FR were our answer. Well, it seems the problems are more complicated than meets the eye. Originally we felt the /FF scheme was an overkill. The /FT switch was supposed to be care-free. We knew from the beginning that /FF would work, of course. Still, our attitude in programming is a lot more exact. Once we allow ourselves to go easy way and the sloppiness to solve problems, you will end up having a tool whose behavior is unpredictable at best. That is not how we do things. (That is why my writing and documentation always plunge into a deep pile of gory details.) This "Fuzzy" solution to solve the timestamp problem is not an ideal solution in our mind, albeit the problem at hand seems quite elusive. The reason for our failure to deliver a complete solution was due to our lack of understanding on exactly how various timestamp- related problems were encountered by XXCOPY users. So, keep in mind when you report a problem, the more detailed observations you make, the better the solution will be. ------------------------------------------------------------- So, please make it just plain /FF and see what happens. If it does not work with the most common setting of NTFS vs FAT, we are willing to release a special "debug" version to generate some debugging info to analyze this case further. To help you test the /FF switch, I recommend the following command which XXCOPY \src\ \dst\ /LDT /DX /FF2 You should choose a pair of directory which are similar (such as a directory and its backup counterpart). Here /LST makes the operation "list-only" without carrying out the copy operation (and displays the Date and time). /DX selects files with that are not the same (ignoring the file size --- /BI would look at file size). If you don't add the last switch (FF2), it shows the files that are different in time in any way. With /FF2, the files that are within 2 seconds will not be shown in the list. By increasing the fuzzy range (say /FF24H), you can reduce the number of files listed as different. (the larger the fuzzy range you make, the greater the number of files become treated as having the same timestamp). Anyway, this experiment leads us to come up with yet another way to create a novel file-filtering scheme... That's enough for now. Kan ============================================================ At 2001-06-22 10:36, you wrote: >Ken, > >You da MAN! the /FF switch works perfectly for my application (synching >between NTFS and FAT32. I made it //ff60 and it's terrific!
This message if part of XXCOPY's message Archive. The archive contains all the messages posted at Yahoo!Groups: XXCOPY.