On Sun, 5 Aug 2001, Stefan Ulrich wrote:

> >> I must admit that from reading RFCs 1738 and 2396, I couldn't tell
> >> for sure whether the above URL is correct or whether it should be
> >> file://localhost//path/to/file
>
> > Both are valid URLs, but the latter example is an absolute path while the
> > first one is a relative path.
>
> sounds reasonable; but what's strange is that libwww (and also browsers
> like netscape, lynx or mozilla) *all* seem to treat an url-path with one
> `/' after `localhost' as absolute, not as relative path.

Thanks for your reports, details and patience!

It is indeed a clear indication that I am pretty lonely in my interpretation
of how to deal with the file: URLs without an initial slash. They all seem to
assume that if the "relative" paths are all specified from the root /.

(I also just noticed that libcurl treats file://localhost/tmp/data and
file:///tmp/data differently, which I figure it shouldn't.)

> > How does libwww treat the path in 'file://localhost/foo' ? as '/foo' ?
>
> Yup;

So then I suggest the attached patch to make libcurl act like "the rest". It
makes all these URLs perfectly valid and refer to the same file:

        file://localhost/foo
        file://localhost//foo
        file:///foo
        file:////foo
        file://whatever/foo

Does that fit your view and the other tools' views of reality?

-- 
     Daniel Stenberg -- curl dude -- http://curl.haxx.se/