On Tue, Dec 23, 2025 at 13:41:07 +0100, Nicolas George wrote: > [email protected] (HE12025-12-23): > > The buffer cache is actually pretty good at such things. You won't hit > > disk too often > > It will not *read back* from the disk too often, but it will be > *written* to disk very frequently; tmpfs avoids that.
I have absolutely no idea what Albretch is trying to do. So I'll just address this tangent instead. If you're using a temp file in a shell script, typically that temp file is very short-lived. Almost always less than one second. A temp file that's created, written to, read from, and deleted all within such a short span of time will probably *not* be written to a physical device. As long as you didn't do something silly like turning on an "always flush" option in the file system mount options, and you aren't using a network file system, the data are highly unlikely to land on a disk. There might be a write to update the *directory* metadata (mtime), since the directory is modified once when creating a new file, and once when unlinking the file. But that should be a quick, one-time operation long after the file has been deleted. It might even be postponed long enough that some *other* temp file in the same directory will cause a different mtime update instead. Of course, using a tmpfs ensures that there will never be a physical write, even for mtime updates. To the OP: please tell us WHAT YOU ARE TRYING TO DO. Specifically.
