klamer@mi.eltn.utwente.nl (Klamer Schutte) (06/03/91)
In <1991Jun1.091604.8465@crash.cts.com> cwr@pnet01.cts.com (Will Rose) writes: >There was also something strange about the tar file of that posting - >the files unpacked totalled ~25K, while the archive itself was ~40K >(from memory). I had another archive recently that showed a similare >effect, tho' not so marked. Is this a 'feature' of minix tar? I don't >think all tars have that sort of overhead. I don't know about the tar file -- but do know the (minix) tar behaviour. In general will a tar file which consist of small files and/or with very much directories give a big overhead. This because of: 1) Each file has a 512 bytes header. 2) Each file will be (zero) padded to a 512 bytes boundary 3) Minix tar (each POSIX compliant one) will for every directory also include a 512 byte header. Older tar (e.g. the SunOs one) do not do this. 4) Tar terminates the tar file with an empty 512 bytes block. For the file aap/noot/mies/wim, length 2 bytes this will give: aap, noot, mies : directory == 3 * 512 = 1536 bytes wim : file == 1 * 512 + ((2+511)/512) * 512 = 1024 bytes trailer: 512 bytes. Total length will thus be 3k, with only 2 bytes of data. This is quite an overhead. Traditinal tar's (minix-tar before 1.5) will give only 1.5k. Klamer PS But remember that all those blocks with zeros can be compressed very effective. tar+compress is about as efficient as shar+compress. -- Klamer Schutte Tel: +31-53-892786 Fax: +31-53-340045 Faculty of electrical engineering -- University of Twente, The Netherlands preferred: klamer@mi.eltn.utwente.nl SMTP: klamer@utelmi01.el.utwente.nl