On Fri, 18 Jan 2019 03:13:05 -0500, Paul wrote:

It's when file system operations are outstanding, and you
pull the plug, that corruption of the last operations can
occur.

Hi Paul,
Thanks for that information - which - will help everyone who cares.
I don't doubt anything you say - as - I have been burned empirically.

It's actually embarrassing to admit, but - in the past - about in the days
of 100GB and 500GB USB drives being "all the rage", I've "corrupted" entire
USB drives more than once - where - while I don't doubt what you say - it
isn't always that simple.

It "might" be that simple - but - I was "sure" I waited for file transfers
to end, and _still_, entire USB drives lost their minds on me.

When I "recovered" the data, it was flat (i.e., no folder hierarchy) and
all the files lost their first letter, where, for example, if the file name
was "foo.txt", it turned into "oo.txt" (as I recall) or maybe "Xoo.txt" (I
don't remember - I just remember the first character was gone).

Using a "safely remove" does a flush() that transfers outstanding
write operations to storage. Once the cache is flushed, you
can pull the plug.

Yup. That's my point to the OP.

If the OP opts for a USB drive, particularly on large file xfers, then a
"safe" return to earth is something the OP _must_ be cognizant of.

A number of file systems (NTFS and EXT4 for example),
are journaled, and even if you pull the plug, at
next mount, the corrupted or half-finished transactions
can be removed. And that leaves the vast majority of the
store intact.

This is a good point, where, for example, the power reliability here in the
environs of the Silicon Valley is worse than that in equatorial Africa
(e.g., we lost power for 18 hours yesterday in a windstorm where we lose
power about once a month). We _all_ have our own power generation units,
but that doesn't stop the PCs from rebooting (yes, I know, there are ways
to ameliorate that).

The main point was to the OP to beware of USB drives - that's all.

If you've managed to disable write caching, then all operations
should go immediately to disk, increasing the chances that
if the LED has stopped flashing, pulling out the cord will
not damage anything.

Nice to know. Me? I avoid USB file transfers like I avoid the plague.
That solves my problem (as my internal disks are large enough).

We manage to use floppies in Windows without too much trouble.
We don't dismount them. There's no cache that's visible. We
wait until the LED stops flashing (or we happen to know no
write operations have been done), and can pop out the floppy
when the thing is quiet. Other operating systems keep trays
or drawers locked to prevent random removal, as another
way to enforce good practices.

This is a good point that, atavistically, the problem has existed.

In Unix, issuing several sync() commands would cause flushing
of content to actual storage, in an attempt to eliminate
damage that could happen soon after that.

Oooooh. I remember some of those. I do. I do. It seemed crazy to do the
same thing twice (literally); but I remember (way back when)

Paul ... you bring back memories ... of days long gone by (I hope)...