tarix is a program for indexing tar archives so that selective extraction can be done rapidly, especially on slow devices like tape drives, providing that you can seek to arbitrary locations (at least on 512 byte boundaries) within the tar archive.
tarix is designed to work as a compression filter for tar with the --use-compress-program option. At this time, it does not actually do any compression, but merely scans the tar archive as it passes through and writes out an index file. Most compression formats do not allow easy, arbitrary seeking within them, so it is not planned to have tarix support invoking a compression program in the near future. Gzip and bzip2 appear to have some ability for seeking within compressed files, but support for this is not planned for the 1.0 release; 2.0 will head down that path.
Because of how tar invokes compress programs, options are generally accepted through environment variables. For manual usage, however, tarix can accept normal command line options.
tarix's output format is geared for simplicity and ease of use in a recovery mode. For those reasons, it is a simple text format. The goal behind this is to allow tarix's indexes to be used from even a basic recovery disk. Using tarix's indexes requires only grep, sed, dd, and tar for basic usage. tarix itself can use the index for fast extraction without convoluted usage of grep, sed, and dd.
Currently tarix targets GNU tar archives. It does currently try and support any of the fancier GNU tar extensions, such as sparse files and such, although it may work with them anyways, since it generally only looks at the file headers within the archive. It can handle GNU's format for very long file names, though it will probably barf on filenames more than 511 chars long. It should also be able to handle standard POSIX archives.
Do not put newlines in your filenames, it WILL break tarix's output format (and probably lots of other things too).
It is unknown if it will work with other tar programs, but the author would be very interested to hear about success or failure with them. If supporting other formats does not excessively complicate tarix, it will be added if and when the necessary information about the format differences is provided to the author.
tarix's format uses 32 bit unsigned integers to count 512 byte blocks, so it should be able to handle archives up to 2 terabytes.
Web Page: http://sf.net/projects/xtar/
Alternate download site: http://reprehensible.net/~cheetah/tarix/