Maybe the gzip algorithm uses some randomness in its compression? No. There's a timestamp in the gzip format (so a file can have its mtime restored when it is gunzipped).
> Of course, I could. But I think this may not be an efficient way if the
> .gz file is too large, say of the order of GB.
The suggestion was to use fifos (named pipes). Thus the uncompressed
versions do not actually exist on disk.
