He expects http://a.b.c/d/../e/ to be the same as http://a.b.c/e/.

use strict;
use warnings;

use Test::More tests => 1;

use URI qw( );

my $base1 = URI->new("http://a.b.c/d/../e/");
my $base2 = URI->new_abs("../e/", URI->new("http://a.b.c/d/"));
my $uri   = URI->new("http://a.b.c/y");

is($uri->rel($base1), $uri->rel($base2));

1;
[download]

1..1
not ok 1
#   Failed test at a.pl line 12.
#          got: '../../../y'
#     expected: '../y'
# Looks like you failed 1 test of 1.
[download]

I don't see anything in http://cpansearch.perl.org/src/GAAS/URI-1.58/t/abs.t that covers this case.

The equality definitely doesn't hold up in file systems (due to links both soft and hard), but ../../../y is much less likely to be correct.

Update: Improved test a bit.

He expects http://a.b.c/d/../e/ to be the same as http://a.b.c/e/.

If I read section 5.2 of RFC 3986 correctly, the OP quite rightly expects that to be.

There is a violation of 5.2 elsewhere, though.

use strict;
use warnings;

use Test::More tests => 1;

use URI qw( );

# if defined(R.scheme) then
#    T.scheme    = R.scheme;
#    T.authority = R.authority;
#    T.path      = remove_dot_segments(R.path);
#    T.query     = R.query;

is( URI->new_abs('http://a.b.c/d/../e/', 'http://a.b.c/'), 'http://a.b
+.c/e/' );

1;
[download]

1..1
not ok 1
#   Failed test at a.pl line 8.
#          got: 'http://a.b.c/d/../e/'
#     expected: 'http://a.b.c/e/'
# Looks like you failed 1 test of 1.
[download]

I would except cannonical to implement 6.2.2.3. It does implement other similar rules.

use strict;
use warnings;

use Test::More tests => 6;

use URI qw( );

is( URI->new('hTtP://a.b.c/'       )->canonical, 'http://a.b.c/',   'R
+FC 3986, 6.2.2.1' );

is( URI->new('http://a.b.c/d/../e/')->canonical, 'http://a.b.c/e/', 'R
+FC 3986, 6.2.2.3' );

is( URI->new('http://a.b.c'        )->canonical, 'http://a.b.c/',   'R
+FC 3986, 6.2.3' );
is( URI->new('http://a.b.c:'       )->canonical, 'http://a.b.c/',   'R
+FC 3986, 6.2.3' );
is( URI->new('http://a.b.c:/'      )->canonical, 'http://a.b.c/',   'R
+FC 3986, 6.2.3' );
is( URI->new('http://a.b.c:80/'    )->canonical, 'http://a.b.c/',   'R
+FC 3986, 6.2.3' );

1;
[download]

1..6
ok 1 - RFC 3986, 6.2.2.1
not ok 2 - RFC 3986, 6.2.2.3
#   Failed test 'RFC 3986, 6.2.2.3'
#   at a.pl line 10.
#          got: 'http://a.b.c/d/../e/'
#     expected: 'http://a.b.c/e/'
ok 3 - RFC 3986, 6.2.3
ok 4 - RFC 3986, 6.2.3
ok 5 - RFC 3986, 6.2.3
ok 6 - RFC 3986, 6.2.3
# Looks like you failed 1 test of 6.
[download]

5.2 describes how to make an absolute uri from a (possibly) relative URL.

In that context, "/d/../e/" is equivalent to "/e/", but the module already does that correctly.

Update: 6.2.2.3 seems more relevant here.

#!/usr/bin/perl --
use strict;
use warnings;
use URI;

Main( @ARGV );
exit( 0 );

sub Main {
    ABT();
}

sub ABT {
    require Test::More;
    Test::More->import(qw' no_plan ');

    my $base = URI->new("http://a/d/../e/");
    my $uri = URI->new("http://a/y");

    is( $base, ( "http://a/d/../e/" ) x 2 );
    is( $base->path, ( "/d/../e/" ) x 2 );
    is( $uri, ( "http://a/y" ) x 2 );
    is( $uri->path, ( "/y" ) x 2 );

    diag("should the following be equivalent??");

    is( $uri->rel($base), "??", q~$uri->rel($base)~ );
    is( URI->new( $uri->path )->rel( $base->path ),
        $uri->rel($base), q~URI->new( $uri->path )->rel( $base->path )
+~ );
    is( URI->new("/y")->rel("/d/../e/"),
        $uri->rel($base), q~URI->new("/y")->rel("/d/../e/") ~ );
}
__END__
ok 1 - http://a/d/../e/
ok 2 - /d/../e/
ok 3 - http://a/y
ok 4 - /y
# should the following be equivalent??
not ok 5 - $uri->rel($base)
#   Failed test '$uri->rel($base)'
#   at - line 27.
#          got: '../../../y'
#     expected: '??'
not ok 6 - URI->new( $uri->path )->rel( $base->path )
#   Failed test 'URI->new( $uri->path )->rel( $base->path )'
#   at - line 28.
#          got: '/y'
#     expected: '../../../y'
not ok 7 - URI->new("/y")->rel("/d/../e/")
#   Failed test 'URI->new("/y")->rel("/d/../e/") '
#   at - line 30.
#          got: '/y'
#     expected: '../../../y'
1..7
# Looks like you failed 3 tests of 7.
[download]