Re: proposal: probe-file -- major behavior change

[Raymond T. <to...@rt...>]

>>>>> "Sam" == Sam Steingold <sd...@gn...> writes:

    >> * In message <4n6...@rt...>
    >> * On the subject of "Re: proposal: probe-file -- major behavior change"
    >> * Sent on 26 Feb 2002 12:18:36 -0500
    >> * Honorable Raymond Toy <to...@rt...> writes:
    >> 
    >> >>>>> "Sam" == Sam Steingold <sd...@gn...> writes:
    >> 
    Sam> OTOH, CMUCL (probe-file "/etc") ==> #p"/etc" is obviously wrong (there
    Sam> is no such file in #p"/").
    >> 
    >> I don't think that this is so wrong.  Some OSs (Solaris) still
    >> actually let you read a directory as a file.  Well, at least "cat
    >> /etc" produces something.

    Sam> In CL, there is a big difference between a file and a directory.

But a file is 

        a named entry in a file system, having an
        implementation-defined nature.

so I'm not sure what the right answer is.  I can't find the glossary
entry for directory.

    >> So maybe someone really wants to know that /etc exists?

    Sam> so how is the user supposed to find out whether /etc is a file or
    Sam> directory?
    Sam> both CLISP and CMUCL have non-portable ways to do that.
    Sam> I propose to make PROBE-FILE useable for that.

I don't know.  If CL doesn't specify a way to tell if /etc is a file
or directory, whatever you do is non-portable, so you might as well
make it non-portable.

Ray