-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Wed, Nov 18, 2015 at 08:54:46AM -0600, David Wright wrote: > On Wed 18 Nov 2015 at 08:47:59 (+0100), to...@tuxteam.de wrote: > > > Here, gunzip is seeing the file name. Since gzip, by convention, removes > > the original (and gunzip the compressed), effectively replacing each by > > the other), those programs are extra careful. The file name pattern is > > part of the interface specification. > > [...] > > > And that makes sense. In a way yes, it has two personalities: an > > "in-place" (de) compressor and a filter. The first one enforces a > > file pattern convention because it's (potentially) deleting stuff. > > > > Not that I personally like the in-place thing, mind you. I guess > > it was born out of necessity: if you are zipping a file which is > > much bigger than what you've space left on device, this interface > > makes sense. > > What you appear to be saying here is that the new version overwrites > the head of the file while the program is still reading the tail. > This doesn't happen: a new file is written and, when complete, the > old file is removed. So there's always a period when both compressed > and uncompressed versions exist simultaneously on the filesystem.
What I was saying was that the interface *allows it*. Whether it's implemented this way is another question, but... watch this space. regards - -- tomás -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.12 (GNU/Linux) iEYEARECAAYFAlZMjmQACgkQBcgs9XrR2kaaWwCfcGB2Ab4/eQnGWaqDl32xezkX 41sAn3z3/uQVzBsBX0EhzINKm7hnh2Ws =kCA8 -----END PGP SIGNATURE-----
