[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

a bug relating to carry-tapes and a question relating copy file



    Date: Tue, 15 Sep 87 11:08 EDT
    From: John Watkins <jwatkins@STONY-BROOK.SCRC.Symbolics.COM>
        Date: Fri, 4 Sep 87 09:59 PDT
        From: Siskind.pa@Xerox.COM
        1) As of 7.1 the carry-tape system no longer defaults to device "cart"
        so when I issue a carry-dump or a carry-list it gives an error which
        I must proceed from and call up a menu to set the device. Further,
        it doesn't remember the setting so I must do this every time.
        This is very annoying and should be fixed. 

    I agree.
                                                   Either provide a way
        that this info can be passed as an argument to tape:carry-dump etc.
        or a way I can set the default myself.

Apart from the problem that I can't run the carry tape system without
going through the error handler to set the drive, another inconsistency
is that tape:carry-load takes the keyword parameter :host while
tape:carry-dump and tape:carry-list call that parameter :tape-host.

                                                Also when writing a carry tape
        that takes more than one tape, if an error ocurs while reading back
        the a tape, all further tapes can not be read. This is a bad bug
        which bit me and makes the carry tape system very unrobust.

You did not comment about the above problem.

        2) How can I copy a whole tree of directories from a LMFS to a
        UNIX 4.2 with one command. That is what is the equivalent to
        the ** convention for the UNIX pathname parser? I know that
        UNIX itself has no convention but at least the lispm pathname
        parser should handle it and allow me to do a copy multiple files
        and diorectories with something like
        :copy file bd:>Siskind>**>*.*.* marxc:/user/siskind/**/*.*.*
        Also it would be nice to have a way to optionally preserve
        version number when going to UNIX. Many sites do not have a LMFS
        machine directly on an external net or only have tape drives
        on a UNIX machine so transfering files often has to go through
        some UNIX intermediary. I would like to be able to preserve
        my verison numbers in such a case.

    Here I have a problem. I don't see why we should implement
    functionalitly that the designers of UNIX didn't implement. But I will
    forwardd your suggestion along. After all I'm just one voice. 

I don't agree, at least with respect to the ** convention. The issue is not
whether UNIX implements this feature. It is a capability of the lispm general
pathname and file capability. All that is needed is to determine some desirable
syntax and augment the UNIX pathname parser on the lispm to generate :wild-inferiors
when appropriate. The rest should happen autoamtically.

In general a really bad problem with carry tapes is that I cant write one from
one directory and load it up on another directory without redirecting practically
every file by hand. I want to be able to dump all of my files from MIT,
where my directory is z:>Qobi> and bring them up at Xerox PARC, where my
directory is bd:>Siskind>. There is no easy way to do this with either carry tapes,
distribution tapes, or the backup dumper. And I would like to preserve version
numbers in the process. It would be nice if I could go through a UNIX machine if
needed, as the cart tape mechanism is very unreliable, at least for carry tape
format.  THIS IS A SORELY LACKING FEATURE OF THE SYMBOLICS ENVIRONMENT!

    If you need it badly, look at fs:directory-list. It returns a consed
    list of of filenames. On element of the filenames attributes is
    :DIRECTORY T. You could write a function to check for this and if
    present recurse on the director to get subdirectories.

        Please respond to Qobi@XX.LCS.MIT.EDU as I am leaving Xerox
        today and my account will be closed down. (The process of
        transfering files between MIT and Xerox has raised the
        above issues.)
                Jeff
        -------