Lessons of resurrecting literary files from my late ex-wife's Digital Audio G4

[DavidHertzberg]

A long-existing facility of Retrospect that I'd never used turned out to be key to implementing Phase 1 of my project of converting my late ex-wife's 1990–2005 literary and artistic files for future use by any of her friends or relatives.  Phase 1 is the literary files; Phase 2—the artistic files—may require some programming.  Retrospect solved the second of two Phase 1 problems, but IMHO other administrators should also know about the first problem.

These files are on a HDD in a Digital Audio G4 that my ex-wife had given me in 2005 after I copied the HDD contents onto a PowerBook she'd bought.  The literary files were written 1990–2005 using Mac MS Word 5.1a under OS 9.1, so they're in .doc format (no file extension on Classic Mac OS)—which I'm not sure will be readable by a future word processing app.  When she complained in 2015 that she couldn't open the old files in Mac Word 2011, I dragged her G4 out of the back of my bedroom closet,  took a thumb drive containing several hundred files over to her apartment, and demonstrated that she could with significant difficulty convert them to .docx format.  However I was occupied with  the ultimately-fatal illness of my longtime guitar teacher and friend , so I didn't then volunteer to do the several days of conversion work for her.  She died 13 January, but her executor can't yet get me into her apartment to see if she later did the time-consuming conversion herself—which I doubt she did except for a few selected files.

So four weeks ago I decided to do the conversion myself, using LibreOffice Writer on my 2016 MacBook Pro.  I copied the files in her old "Microsoft Word" HDD folder onto a "Microsoft Word etc." folder on a thumb drive, copied the files from that folder onto a "Converted ..." folder on the same thumb drive, and then plugged the thumb drive into my MBP so I could convert them using LibreOffice Writer.  I was able to convert 800 of the 900 files in either 20 seconds per file if LibreOffice recognized the file as .doc, or 30 seconds per file if it didn't.  Note that the thumb drive, which I pulled out of its Best Buy 2015 wrapper, is formatted for Windows—which turned out to be significant as per two paragraphs down.

The first problem was that, for another ~40 filenames, LibreOffice wouldn't do a name-preserving Save As.

In 2015 I had encountered a filename for which Word 2004 replaced everything after "#2." as the the first three characters with "docx'".  I then though this was a Microsoft stupidity, but 3 weeks ago I realized that the stupidity is a consequence of a Windows-friendly refinement in the way Apple handled the transition from Classic Mac OS to OS X.  Classic Mac OS prohibited only the character ':' in filenames, but Windows NTFS prohibits basically all the characters listed in this Wikipedia article section (note '.' is specially treated). Apple's refinement—to avoid making users go through an editing process while upgrading—was to allow those characters in existing filenames, but to require that they be eliminated in every Save As to a Windows-formatted drive.  In that case Apple's Save As code—used in both Microsoft Word 2004 onward and in LibreOffice—assumes the rightmost '.' in any filename is the beginning of a file extension, which wouldn't be true for a filename still in its Classic Mac OS format—which is still OK on OS-X-formatted drives.

That meant I had to pre-replace filename characters banned by by Windows NTFS, of which  '.' and '/' and '?' and straight-double-quote were the most common in my ex-wife's files.  I did that by copying-and-pasting each file whose filename contained such characters from the G4 HDD to the G4 desktop, renaming the desktop copy with '_' or '-' or " quest" replacing its illegal characters, copying the desktop copy to its proper hierarchical place in the "Microsoft Word etc." folder on the thumb drive, and then deleting the desktop copy.  LibreOffice on my MBP can convert a file with such a pre-replaced filename to .docx format with a name-preserving Save As.  This is OK, because the friend/relative may use Windows.

The second problem was that ~100 of the files converted in LibreOffice Writer as simply rows of hash symbols—rather than their contents when I opened them in Word 2004 on the HDD under OS X 10.3 Panther.

I decided OS X 10.3 on the G4 had somehow messed up those files when I copied them from the HDD to the "Microsoft Word etc." folder on the thumb drive, and that this was probably because I'd used the Finder's ability to copy nested folders.  At this point I remembered that Retrospect has long had a facility named Duplicate in the version 6.1 I had on an OS X 10.3 drive on the G4.  Duplicate (still named that in Retrospect Windows, but renamed Copy in Retrospect Mac 8—distinct from Copy Backup or Copy Media Set) copies entire volumes or defined-to-Retrospect subvolumes (renamed Favorite Folders in Retrospect Mac 8 ) between a Source and a Destination.  It goes beyond the Finder in having an option to compare the copied files and show the results in a log, and another option to delete successfully-copied files from the Source.  What  from  fixed my hash row is that Duplicate's directory traversal is independent from that of the Finder.  But Duplicate also logged two files it couldn't copy; I'd not spotted the character '/' in their filenames.

LibreOffice on the MBP successfully converted most of the files I re-copied from the HDD to the thumb drive's "Converted ..." folder using Duplicate on the G4.  Then I Finder-copied one by one the dozen files LibreOffice still couldn't convert.  One file ultimately converted as a garbage line; it was garbage on the G4 HDD.  IMHO the fact that Finder-copy didn't mess up a file when it was copied individually means that there's a bug in hierarchical Finder-copy, at least in OS X 10.3.