Because cross-device copying is the only place where this operation is slow, with modern HDDs all other operations are instantanious for all intents and purposes.
Good idea but you really should have put more tought on it.
going forward it seems reasonable to assume ubiquitous network connectivity, so you can simply treat external drives as a cache and show a non-modal alert "please reconnect for sync, expected to finish within X minutes" iff the amount of data synced is large enough to matter via a 3G connection.
You obviously unmount or decouple your external drive in software first and that's a good place for the operating system to inform you that:
Ongoing I/O for your USB mass storage device.
Wait for completion before unplugging.
And if there are many parallel I/O streams going on and this dialog is invoked, automatically prioritize anything that's being copied to the external hard drive first so that the user can unplug it as soon as possible.You haven't met most users, have you?
I'm at my friends house, copying files from a network share over his wifi.
I'm copying from mounted network shares at home over the internet.
I'm putting files on, or pulling them off, a USB stick.
If copy turned into the feature you described, there would have to be a 2nd "no, REALLY, copy" command. Hardly a step forward.
That being said, there's no reason it couldn't work the way you describe if the system knows it's copying to/from local physical discs.
You do? I'm often too lazy to.
Also, besides copying files to removable drives, etc, the primary reason I copy a file is because I want two separate copies. Otherwise I would move it.
Cue iOS.
Which was a nice experiment which proved that it didn't work. You know what you can't do on iOS which most users want to be able to do?
Yeah. Work with the same document from multiple viewpoints, or applications. You can't work with documents across applications on iOS because you supposedly don't have files.
Like I said: A nice experiment, but it proves that "no files" only works for the simplest of use-cases and limits the ability to work with data beyond the limits that people reasonably expect to find.
What I would like is that copies seem instant and transparent. Actually, I wouldn't like to think of copying at all. I "copy" (or merely just order) a file to (appear in) some place and it immediately becomes accessible from there. Or I decide to move it back a few seconds later and I can do that.
In reality the computer should manage the low-level copying internally and consider it more like syncing.
I shouldn't have to care about how long something takes or how the kernel will queue the actual copy operations, or whether it copies anything at all but just creates a copy-on-write mapping to the same data. I also shouldn't have to wait for copying: I should be able to just copy something and immediately turn off the computer. The operating system would continue copying the blocks, if necessary, when it's restarted later. All copy operations ought to be logically atomic as well: either a copy is immediately made or it isn't, the physical details of copying and remapping the corresponding blocks shouldn't be any of my concern.
So, an immediate, transaction-based and logically coherent file operations with automatic kernel-level delayed-I/O backend is what I would call the 2000's.
I hope that there already is at least one research O/S somewhere that would do this.