Hacker News Reader

Comment by opless | Hacker News Reader

Tar IS seekable, as it was designed especially for tapes.

The modern usage of it adds compression, which you have to undo first before doing your archive operations ;)

Zip was > tar (and cpio) in that regard because you didn't have that two step process and only had to have enough spare disk space for the extracted file, rather than extracted file + decompressed tar

View on HN

Tar is seekable in the literal sense, you have to seek around until you find the file you want. You have to look at each header, if it's not the file you want you skip over that one file and read the next header. Because you only know how long that one file is. So you cannot even know where the third file is without having processed the headers of the previous two. It's "read, skip, read, skip, read skip" until you maybe find the file you wanted. ZIP otoh has the central directory index where you can look things up much faster.

Zip still > tar in that regard, since you can list archive contents (and extract individual files) without having to go through the entire thing every time, hence zip being suitable as a local file system (à la opendocument or ooxml).

Zip is terrible as a local files system, unless you consider read only (randomly, and not many files). Writing/append to a file requires some form of garbage organization/compaction - which requires rewrite of the entire CEN. Unlike the regular files/entries, CEN doesn't have a checksum, so corruption there is pretty bad.

Each file being compressed individually means low compression rates, the deflate overhead might be acceptable for slow mediums (although in that case the memory overhead would be non-trivial).

Reply on news.ycombinator.com