Summary: Copying Cartridge tape SunOS distributions

From: Dana Roode (roode@anansi.acs.uci.edu)
Date: Sat Dec 22 1990 - 12:26:24 CST


On Wed, 19 Dec 90 19:30:47, I sent a query out about copying 1/4"
SunOS distributions (given licensed right to do so). I've gotten
over 40 responses - I thank everyone who took the time to reply.

My conclusions follow:

    - There is no blocking factor on cartridge tape in the sense that
      1/2" magtapes have blocking factors. Cartridge tapes use fixed
      512 byte physical blocksizes, period.

    - Assuming a sufficiently large file on the tape, no matter what
      request size you use on a READ, you will get that many bytes
      back. A tape written with 10 512 byte "blocks" can be read via
      5 1024 byte reads - the data is the same either way.

    - Specifying blocksizes does have some effect on the speed of the
      I/O operations, presumably due to the way the Operating system
      drivers, the SCSI bus and the tape controllers work.

    - 126b (b=512 bytes) is one "recommended" blocksize, one rumor was
      that it was because of some limit on the SCSI bus. I've also
      seen 200b as another "recommendation", but there was a recent
      posting to sun spots where the sender had experimented with large
      blocksizes (1000b) and found them faster. I too have been using
      huge blocksizes with success now. At any rate, the blocksizes
      are pertinent due to system buffering of some sort, not because
      of the physical layout of the tape.

    - There appear to be some software bugs under SunOS 4.0(.x?)
      somewhere in the cartridge tape code. I've seen code written to
      reposition the tape after each file was written via rewinding
      and spacing forward again. For some reason the tape being
      written loses its position. I'd like to hear from anyone that
      can verify this. I just found I had to do this under 4.1, so
      perhaps its more than a mere software bug. Only seems necessary
      sometimes.

    - My odd problem where I have to specify a blocksize of 1b to read
      the first file on a Sun 4.1.1 distribution tape to avoid getting
      an "i/o error" seems to be dependent on the operating system and
      tape drives I use. People at other sites read the same file
      with a large blocksize just fine, as do I on other 4.1 systems.
      I suspect something odd about our installation, or perhaps
      another 4.0.3 bug?

    - A recommended reading: "Tutorial on 1/4" Tapes" from Sun, their
      part number 800-1315-02.

How do people copy distribution tapes?

    - MANY people recommended "copytape", a public domain tape copier
      written by David Hayes of the US Army AI Center. It is available
      from the uunet comp.sources.unix archives (volume 10).

    - Several people mentioned "tcopy", which comes with SunOS 4.0 and
      after.

    - Several people use the Sun Consulting Services "special" called
      "tputil". Tputil works well, and will also verify tapes.

    - Many people do what I do: use scripts that do multiple dd's.

Well, that's enough. I understand from one responder that someone in
Sun Consulting is working on a program/script that will take a
distribution CD and create bootable tapes from it. That sounds like a
good idea.

              Dana Roode
              UC Irvine



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:00 CDT