How You Can Lose a File Despite Three Layers of Backup (and How To Avoid It)

Originally published at: How You Can Lose a File Despite Three Layers of Backup (and How To Avoid It) - TidBITS

You’d think that three separate continuous file-archiving systems would have been enough to protect Glenn Fleishman from his own mistake after modifying a file whose original he wanted to see. Unfortunately, circumstances conspired. His cautionary tale offers a lesson.

2 Likes

I have used a similar manual versioning process over the years, though more infrequently since I retired.

Glenn, do you, or does TidBITS or Take Control have a nice shareable script that facilitates the process?

DEC’s VMS operating system did versioning for all files. It has been so long, I do not remember the file name structure now. Interesting to hear of a manual alternative.

1 Like

If you want to manually retain versions of files, you might also consider a version control package like Git. Although designed for source code, it works well for all kinds of files (but may not be very efficient for compressed/binary file formats).

I use it to maintain an archive of the files on my local web server (which are mostly HTML text). But you have to remember to do a git add/git commit whenever you want to preserve the current version. Nothing is automatic here. But it may be more convenient than keeping every version stored with a different file-name.

1 Like

Thanks for sharing this dreadful tale :sweat_smile:.

I have had a related problem a couple times recently. Somehow I lost or overwrote my newest work on a file. I don’t know how it happened though. But it seems to be related to Dropbox delayed syncing and moving between various devices to edit the same file.

I hate losing work and I did. I’d like to figure out how to reproduce it. But somehow I think Dropbox is destroying my data by getting confused about where the latest copy is.

I’m a big Dropbox fan. But this is a problem. And I’m still pissed that they quietly removed syncing Mac aliases last year. That was not cool and I still don’t have a solution to those handy shortcuts to key folders.

1 Like

All by hand. Intentional, in fact. We want to have the motor memory of having to do it and having done it.

1 Like

I kind of treat Backblaze, Dropbox, and Time Machine as my versioning system. Dropbox in particular is extremely good at this — if I hadn’t missed losing the extended history option, this story would never have been written! But it’s less explicit. I like an automatic versioning system (continuous version backup) and an explicit manual one (using folders and naming).

And loving it, right?

Long ago, we used Subversion for TidBITS articles. Even with the integration with BBEdit and automation with Keyboard Maestro, having to update before working and commit after making changes was a royal pain in the butt. Even worse was when we had an actual conflict and had to reconcile it.

It was a huge relief to switch to Google Docs, which saves so constantly that even when something goes wrong (with Google Docs itself, the browser, or the Mac), I usually lose no more than the last few characters I typed. And because it versions everything automatically, there’s never any problem with going all the way back to the beginning.

Obviously, Glenn wasn’t using Google Docs for this work, but it’s one of the reasons we rely so heavily on it for all TidBITS writing nowadays.

2 Likes

One extra tip: despite my long history with the Mac, I had missed the fact that the Finder sorts alphabetically and numerically when punctuation is involved, meaning you don’t need to add leading zeroes

Glenn,

Yes, the Finder does sort that way, but other tools (such as AppleScripts or Acrobat when working with multiple files) might not. So be careful.

I still insert leading zeroes, from habit as much as keeping everything neat.

Anyone else old enough to remember file versions on VAX/VMS??? That was a really nice feature.

Timemachine deserves a little blame, backup the original file on open if you haven’t seen it before, thanks timemachine.

Wasn’t default versioning also the hope for ZFS on mac, still not clear why we ended up with APFS instead.

Definitely an important article for everyone to read.

As a tech consultant, I’ve been advising clients to keep multiple iterations of documents (preferably with useful suffixes like your SOP) for decades.

It’s the only way to truly cover your ass.

Of course eons ago this sometimes caused space storage issues, but a non-issue once the age of bigger drives came along.

And I’d like to point out an important detail that I would suggest adding to this article: Cloud storage services like iCloud, Dropbox and OneDrive allow for disk-saving modes, where some files aren’t stored locally. Which means many folders and files may only exist in the cloud and not actually on your hard drive. Though this may not apply to a document you modified in the last 45 days, there is the potential for this to bite you in the ass. Why/how? Because the only real version of these documents exists in the cloud, and nowhere on your Mac. This means that neither Time Machine or Backblaze (or any other backup service or utility) will be able to backup these folders or files. So there’s a very real potential for data loss here.

The best solution to this is, if you have the HD capacity, turn off this feature, so that all the files actually exist in their entirety locally.

1 Like

I do sometimes forget how little storage we had in the past!

This is a very smart point to make. I wrote a column (and parts of many others) about the danger of this regarding iCloud Photos over at Macworld. If you choose Optimized Storage on all your devices, the only “truth” is the cloud. So if Apple ever has a double-triple disaster, all your high-res media is on their servers and could be lost. I’ve written about strategies to make sure you download your library locally so you have a local copy of the images!

1 Like

It seems to me it would have been simpler if you had forced an immediate TM backup (Back Up Now from the Apple TM Menu) of a closed file that this file would be instantiated in the TM structure as the base file. All subsequent file close operations would be treated as versions.

BTW - Are you using DropBox’s Sync Online option?

Yes. You got quite a lot of files (each time you save) but it was so much easier to manage your files from the command line. I can’t tell you how many times I went back through versions to retrieve something I needed or to so easily back out of edits. When you work like this, you really depend on it. Drives weren’t as reliable or long-lived, so I made printouts as well.

Unix worked well too for not losing stuff because the user always had mostly full control. The times you do suddenly fully lose an hour or two’s of your own work (or more), you usually can recreate it in perhaps a quarter of the time it took to create it the first time. I think this true on any system.

It’s gotten more important to know how each application works and how the (GUI) OS works. I do not think things have evolved for the better although I realize many do.

If I’m being smart, I generally try to keep the original version of a file, if I didn’t create it.

My fantasy is that every application would incorporate a “diff” function where one could get a listing of changes from two versions of a file, with as much detail as possible. I must have used Unix diff 100,000 times in my life for all types of situations. Sometimes just to get a handle on what was lost. It’s not going to tell you about a destroyed file, but it can be the next best thing.

I’m using GIT https://git-scm.com for local staging and versioning, with the GitFinder GUI (fully integrated in the finder, like, say, Dropbox or iCloud) https://gitfinder.com. I can’t believe that people still add numbers to their files to create versions :slight_smile:
cheers
–e.

Sorry Enrico, but for collaborative work I use a similar manual process to Glenn. Generally I include a date in the filename (eg xxx_29sep2021.docx - this format avoids the crazy mix of date formats around the world but I appreciate it does not help sorting files in Finder) and those who send an edited copy back to me add their initials. Redundant versions and feedback go into an “archive” folder, to be trashed eventually.

It is cumbersome, I know, but years of bad experiences with so-called collaborative systems (starting with Lotus!) taught me it is worth the extra effort.

But keep in mind that Time Machine can take quite a while to do even a small backup. So you would be sitting there waiting for Time Machine to complete 100%, just for the sake of being able to work with one document. Doesn’t seem like an efficient use of time.

Making a copy of the original, before making changes, is the safer approach, as you know definitively that you have a copy of the document before modification.

2 Likes

If you use xxx_2021/09/29.docx, the date should be obvious to anyone no matter their local date format, and Finder will sort all xxx’s together then the individual ones by date.