On Mon, 24 Jan 2000, Patrick wrote: > Thanks. The big q is what's the recovery procedure?
Fixing by hand or writing a tool do to so. But hunt the net first, someone might have done it already. Let's supose your file got corrupted by a ASCII upload from UNIX -> NT. First, you must guarantee that you're downloading the file back to UNIX in BINARY mode so as not to make matters worse. Then you use some clues to get the bytes right again. Sometimes it can be automated, but most of the time not. And you WILL have to write the code to do so. If you uploaded stuff from Unix to M$dog-like systems (NT, Win9x...) in ASCII mode, AFAIK the corruption is that every 0A0D pair in the input stream (the corrupted file) *might* be either 0A0D or 0A only. You could write code that uses heuristics to read in the tar file, use the offsets, file size, headers and CRC (if there is one) from the tar file itself (as long as it does not contain a 0A0D in there... :-( ) to extract the corrupted chunk which should be one file, then apply all possible transformations (reduced by the heuristicts. E.g.: if you have the original file size, you know how many bytes were corrupted, and that reduces the set of possible valid 'fixes') and use the CRC to verify if the 'fixed' file is correct. Text files are trivial to fix. They're not even too badly screwed up, just tar can't find them because the offsets in the tar headers are wrong. A program to repair this kind of damage probably exists already. If your data is very important and this type of data recovery is a bit above your current time/knowledge constrains, you'll find someone who can do that for you, but it may not come cheap. -- "One disk to rule them all, One disk to find them. One disk to bring them all and in the darkness grind them. In the Land of Redmond where the shadows lie." -- The Silicon Valley Tarot Henrique Holschuh
