Portability of Tar Features

Summary

A good reference on different tar formats is the tar(5) manpage from libarchive . Of particular interest are four standards supported by GNU tar: Old v7 tar format (--format=v7), POSIX 1003.1-1988 (ustar) format (--format=ustar), POSIX 1003.1-2001 (pax) format (--format=pax), GNU tar 1.13.x format (--format=gnu). Additionally, whenever applicable the two additional formats supported by star were tested: The old v7 tar format is the format used by the tar command supplied with Unix v7, and apparently a common base for the remaining formats. Its defining features are lack of magic bytes and severe limitations (only regular files, hardlinks and symlinks; pathname up to 99 octets; file size up to 8 GiB; user and group stored numerically). The ustar format extends the v7 format by adding more header fields into unused padding space. It provides magic bytes along with version field, user and group names up to 31 octets, support for more file types and extension of pathname length with 154-octet prefix. Some of the implementations used draft version of ustar format that used a different magic bytes and version. The pax format extends the ustar format by allowing arbitrary attributes to be stored as special archive members before the actual file entry. This provides for unlimited length pathnames, file sizes; unlimited precision timestamps, etc. The defining feature of pax format is that it allows for extensions, assuming that incompatible implementations may write the extended attributes as regular files for user inspection. The GNU tar format is derived from the v7 format separately from POSIX formats. It uses the same magic and version as the pre-POSIX ustar format, and is partially compatible with it. However, whereas ustar provides for extending pathname length, GNU tar includes fields for additional timestamps and some other metadata. It also uses a few additional member types to provide long pathnames and support for multi-volume archives. The star format is the form...