diff --git a/tests/test_url.py b/tests/test_url.py
index 7ac5d092..bfdd5bd6 100644
--- a/tests/test_url.py
+++ b/tests/test_url.py
@@ -4,7 +4,10 @@
 import unittest
 from w3lib.url import (is_url, safe_url_string, safe_download_url,
     url_query_parameter, add_or_replace_parameter, url_query_cleaner,
-    file_uri_to_path, path_to_file_uri, any_to_uri, urljoin_rfc)
+    file_uri_to_path, path_to_file_uri, any_to_uri, urljoin_rfc,
+    canonicalize_url, parse_url)
+from six.moves.urllib.parse import urlparse
+
 
 class UrlTests(unittest.TestCase):
 
@@ -347,6 +350,209 @@ def test_urljoin_rfc_deprecated(self):
         self.assertEqual(jurl, b"http://www.example.com/test")
 
 
+class CanonicalizeUrlTest(unittest.TestCase):
+
+    def test_canonicalize_url(self):
+        # simplest case
+        self.assertEqual(canonicalize_url("http://www.example.com/"),
+                                          "http://www.example.com/")
+
+    def test_return_str(self):
+        assert isinstance(canonicalize_url(u"http://www.example.com"), str)
+        assert isinstance(canonicalize_url(b"http://www.example.com"), str)
+
+    def test_append_missing_path(self):
+        self.assertEqual(canonicalize_url("http://www.example.com"),
+                                          "http://www.example.com/")
+
+    def test_typical_usage(self):
+        self.assertEqual(canonicalize_url("http://www.example.com/do?a=1&b=2&c=3"),
+                                          "http://www.example.com/do?a=1&b=2&c=3")
+        self.assertEqual(canonicalize_url("http://www.example.com/do?c=1&b=2&a=3"),
+                                          "http://www.example.com/do?a=3&b=2&c=1")
+        self.assertEqual(canonicalize_url("http://www.example.com/do?&a=1"),
+                                          "http://www.example.com/do?a=1")
+
+    def test_sorting(self):
+        self.assertEqual(canonicalize_url("http://www.example.com/do?c=3&b=5&b=2&a=50"),
+                                          "http://www.example.com/do?a=50&b=2&b=5&c=3")
+
+    def test_keep_blank_values(self):
+        self.assertEqual(canonicalize_url("http://www.example.com/do?b=&a=2", keep_blank_values=False),
+                                          "http://www.example.com/do?a=2")
+        self.assertEqual(canonicalize_url("http://www.example.com/do?b=&a=2"),
+                                          "http://www.example.com/do?a=2&b=")
+        self.assertEqual(canonicalize_url("http://www.example.com/do?b=&c&a=2", keep_blank_values=False),
+                                          "http://www.example.com/do?a=2")
+        self.assertEqual(canonicalize_url("http://www.example.com/do?b=&c&a=2"),
+                                          "http://www.example.com/do?a=2&b=&c=")
+
+        self.assertEqual(canonicalize_url(u'http://www.example.com/do?1750,4'),
+                                           'http://www.example.com/do?1750%2C4=')
+
+    def test_spaces(self):
+        self.assertEqual(canonicalize_url("http://www.example.com/do?q=a space&a=1"),
+                                          "http://www.example.com/do?a=1&q=a+space")
+        self.assertEqual(canonicalize_url("http://www.example.com/do?q=a+space&a=1"),
+                                          "http://www.example.com/do?a=1&q=a+space")
+        self.assertEqual(canonicalize_url("http://www.example.com/do?q=a%20space&a=1"),
+                                          "http://www.example.com/do?a=1&q=a+space")
+
+    def test_canonicalize_url_unicode_path(self):
+        self.assertEqual(canonicalize_url(u"http://www.example.com/résumé"),
+                                          "http://www.example.com/r%C3%A9sum%C3%A9")
+
+    def test_canonicalize_url_unicode_query_string(self):
+        # default encoding for path and query is UTF-8
+        self.assertEqual(canonicalize_url(u"http://www.example.com/résumé?q=résumé"),
+                                          "http://www.example.com/r%C3%A9sum%C3%A9?q=r%C3%A9sum%C3%A9")
+
+        # passed encoding will affect query string
+        self.assertEqual(canonicalize_url(u"http://www.example.com/résumé?q=résumé", encoding='latin1'),
+                                          "http://www.example.com/r%C3%A9sum%C3%A9?q=r%E9sum%E9")
+
+        self.assertEqual(canonicalize_url(u"http://www.example.com/résumé?country=Россия", encoding='cp1251'),
+                                          "http://www.example.com/r%C3%A9sum%C3%A9?country=%D0%EE%F1%F1%E8%FF")
+
+    def test_canonicalize_url_unicode_query_string_wrong_encoding(self):
+        # trying to encode with wrong encoding
+        # fallback to UTF-8
+        self.assertEqual(canonicalize_url(u"http://www.example.com/résumé?currency=€", encoding='latin1'),
+                                          "http://www.example.com/r%C3%A9sum%C3%A9?currency=%E2%82%AC")
+
+        self.assertEqual(canonicalize_url(u"http://www.example.com/résumé?country=Россия", encoding='latin1'),
+                                          "http://www.example.com/r%C3%A9sum%C3%A9?country=%D0%A0%D0%BE%D1%81%D1%81%D0%B8%D1%8F")
+
+    def test_normalize_percent_encoding_in_paths(self):
+        self.assertEqual(canonicalize_url("http://www.example.com/r%c3%a9sum%c3%a9"),
+                                          "http://www.example.com/r%C3%A9sum%C3%A9")
+
+        # non-UTF8 encoded sequences: they should be kept untouched, only upper-cased
+        # 'latin1'-encoded sequence in path
+        self.assertEqual(canonicalize_url("http://www.example.com/a%a3do"),
+                                          "http://www.example.com/a%A3do")
+
+        # 'latin1'-encoded path, UTF-8 encoded query string
+        self.assertEqual(canonicalize_url("http://www.example.com/a%a3do?q=r%c3%a9sum%c3%a9"),
+                                          "http://www.example.com/a%A3do?q=r%C3%A9sum%C3%A9")
+
+        # 'latin1'-encoded path and query string
+        self.assertEqual(canonicalize_url("http://www.example.com/a%a3do?q=r%e9sum%e9"),
+                                          "http://www.example.com/a%A3do?q=r%E9sum%E9")
+
+    def test_normalize_percent_encoding_in_query_arguments(self):
+        self.assertEqual(canonicalize_url("http://www.example.com/do?k=b%a3"),
+                                          "http://www.example.com/do?k=b%A3")
+
+        self.assertEqual(canonicalize_url("http://www.example.com/do?k=r%c3%a9sum%c3%a9"),
+                                          "http://www.example.com/do?k=r%C3%A9sum%C3%A9")
+
+    def test_non_ascii_percent_encoding_in_paths(self):
+        self.assertEqual(canonicalize_url("http://www.example.com/a do?a=1"),
+                                          "http://www.example.com/a%20do?a=1"),
+        self.assertEqual(canonicalize_url("http://www.example.com/a %20do?a=1"),
+                                          "http://www.example.com/a%20%20do?a=1"),
+        self.assertEqual(canonicalize_url(u"http://www.example.com/a do£.html?a=1"),
+                                          "http://www.example.com/a%20do%C2%A3.html?a=1")
+        self.assertEqual(canonicalize_url(b"http://www.example.com/a do\xc2\xa3.html?a=1"),
+                                          "http://www.example.com/a%20do%C2%A3.html?a=1")
+
+    def test_non_ascii_percent_encoding_in_query_arguments(self):
+        self.assertEqual(canonicalize_url(u"http://www.example.com/do?price=£500&a=5&z=3"),
+                                          u"http://www.example.com/do?a=5&price=%C2%A3500&z=3")
+        self.assertEqual(canonicalize_url(b"http://www.example.com/do?price=\xc2\xa3500&a=5&z=3"),
+                                          "http://www.example.com/do?a=5&price=%C2%A3500&z=3")
+        self.assertEqual(canonicalize_url(b"http://www.example.com/do?price(\xc2\xa3)=500&a=1"),
+                                          "http://www.example.com/do?a=1&price%28%C2%A3%29=500")
+
+    def test_urls_with_auth_and_ports(self):
+        self.assertEqual(canonicalize_url(u"http://user:pass@www.example.com:81/do?now=1"),
+                                          u"http://user:pass@www.example.com:81/do?now=1")
+
+    def test_remove_fragments(self):
+        self.assertEqual(canonicalize_url(u"http://user:pass@www.example.com/do?a=1#frag"),
+                                          u"http://user:pass@www.example.com/do?a=1")
+        self.assertEqual(canonicalize_url(u"http://user:pass@www.example.com/do?a=1#frag", keep_fragments=True),
+                                          u"http://user:pass@www.example.com/do?a=1#frag")
+
+    def test_dont_convert_safe_characters(self):
+        # dont convert safe characters to percent encoding representation
+        self.assertEqual(canonicalize_url(
+            "http://www.simplybedrooms.com/White-Bedroom-Furniture/Bedroom-Mirror:-Josephine-Cheval-Mirror.html"),
+            "http://www.simplybedrooms.com/White-Bedroom-Furniture/Bedroom-Mirror:-Josephine-Cheval-Mirror.html")
+
+    def test_safe_characters_unicode(self):
+        # urllib.quote uses a mapping cache of encoded characters. when parsing
+        # an already percent-encoded url, it will fail if that url was not
+        # percent-encoded as utf-8, that's why canonicalize_url must always
+        # convert the urls to string. the following test asserts that
+        # functionality.
+        self.assertEqual(canonicalize_url(u'http://www.example.com/caf%E9-con-leche.htm'),
+                                           'http://www.example.com/caf%E9-con-leche.htm')
+
+    def test_domains_are_case_insensitive(self):
+        self.assertEqual(canonicalize_url("http://www.EXAMPLE.com/"),
+                                          "http://www.example.com/")
+
+    def test_canonicalize_idns(self):
+        self.assertEqual(canonicalize_url(u'http://www.bücher.de?q=bücher'),
+                                           'http://www.xn--bcher-kva.de/?q=b%C3%BCcher')
+        # Japanese (+ reordering query parameters)
+        self.assertEqual(canonicalize_url(u'http://はじめよう.みんな/?query=サ&maxResults=5'),
+                                           'http://xn--p8j9a0d9c9a.xn--q9jyb4c/?maxResults=5&query=%E3%82%B5')
+
+    def test_quoted_slash_and_question_sign(self):
+        self.assertEqual(canonicalize_url("http://foo.com/AC%2FDC+rocks%3f/?yeah=1"),
+                         "http://foo.com/AC%2FDC+rocks%3F/?yeah=1")
+        self.assertEqual(canonicalize_url("http://foo.com/AC%2FDC/"),
+                         "http://foo.com/AC%2FDC/")
+
+    def test_canonicalize_urlparsed(self):
+        # canonicalize_url() can be passed an already urlparse'd URL
+        self.assertEqual(canonicalize_url(urlparse(u"http://www.example.com/résumé?q=résumé")),
+                                          "http://www.example.com/r%C3%A9sum%C3%A9?q=r%C3%A9sum%C3%A9")
+        self.assertEqual(canonicalize_url(urlparse('http://www.example.com/caf%e9-con-leche.htm')),
+                                          'http://www.example.com/caf%E9-con-leche.htm')
+        self.assertEqual(canonicalize_url(urlparse("http://www.example.com/a%a3do?q=r%c3%a9sum%c3%a9")),
+                                          "http://www.example.com/a%A3do?q=r%C3%A9sum%C3%A9")
+
+    def test_canonicalize_parse_url(self):
+        # parse_url() wraps urlparse and is used in link extractors
+        self.assertEqual(canonicalize_url(parse_url(u"http://www.example.com/résumé?q=résumé")),
+                                          "http://www.example.com/r%C3%A9sum%C3%A9?q=r%C3%A9sum%C3%A9")
+        self.assertEqual(canonicalize_url(parse_url('http://www.example.com/caf%e9-con-leche.htm')),
+                                          'http://www.example.com/caf%E9-con-leche.htm')
+        self.assertEqual(canonicalize_url(parse_url("http://www.example.com/a%a3do?q=r%c3%a9sum%c3%a9")),
+                                          "http://www.example.com/a%A3do?q=r%C3%A9sum%C3%A9")
+
+    def test_canonicalize_url_idempotence(self):
+        for url, enc in [(u'http://www.bücher.de/résumé?q=résumé', 'utf8'),
+                         (u'http://www.example.com/résumé?q=résumé', 'latin1'),
+                         (u'http://www.example.com/résumé?country=Россия', 'cp1251'),
+                         (u'http://はじめよう.みんな/?query=サ&maxResults=5', 'iso2022jp')]:
+            canonicalized = canonicalize_url(url, encoding=enc)
+
+            # if we canonicalize again, we ge the same result
+            self.assertEqual(canonicalize_url(canonicalized, encoding=enc), canonicalized)
+
+            # without encoding, already canonicalized URL is canonicalized identically
+            self.assertEqual(canonicalize_url(canonicalized), canonicalized)
+
+    def test_canonicalize_url_idna_exceptions(self):
+        # missing DNS label
+        self.assertEqual(
+            canonicalize_url(u"http://.example.com/résumé?q=résumé"),
+            "http://.example.com/r%C3%A9sum%C3%A9?q=r%C3%A9sum%C3%A9")
+
+        # DNS label too long
+        self.assertEqual(
+            canonicalize_url(
+                u"http://www.{label}.com/résumé?q=résumé".format(
+                    label=u"example"*11)),
+            "http://www.{label}.com/r%C3%A9sum%C3%A9?q=r%C3%A9sum%C3%A9".format(
+                    label=u"example"*11))
+
+
 if __name__ == "__main__":
     unittest.main()
 
diff --git a/w3lib/url.py b/w3lib/url.py
index c3d84664..1bd8c7e5 100644
--- a/w3lib/url.py
+++ b/w3lib/url.py
@@ -10,7 +10,8 @@
 import six
 from six.moves.urllib.parse import (urljoin, urlsplit, urlunsplit,
                                     urldefrag, urlencode, urlparse,
-                                    quote, parse_qs, parse_qsl)
+                                    quote, parse_qs, parse_qsl,
+                                    ParseResult, unquote, urlunparse)
 from six.moves.urllib.request import pathname2url, url2pathname
 from w3lib.util import to_bytes, to_native_str, to_unicode
 
@@ -279,3 +280,174 @@ def any_to_uri(uri_or_path):
 
            # this last one is deprecated ; include it to be on the safe side
            "urljoin_rfc"]
+
+
+def _safe_ParseResult(parts, encoding='utf8', path_encoding='utf8'):
+    # IDNA encoding can fail for too long labels (>63 characters)
+    # or missing labels (e.g. http://.example.com)
+    try:
+        netloc = parts.netloc.encode('idna')
+    except UnicodeError:
+        netloc = parts.netloc
+
+    return (
+        to_native_str(parts.scheme),
+        to_native_str(netloc),
+
+        # default encoding for path component SHOULD be UTF-8
+        quote(to_bytes(parts.path, path_encoding), _safe_chars),
+        quote(to_bytes(parts.params, path_encoding), _safe_chars),
+
+        # encoding of query and fragment follows page encoding
+        # or form-charset (if known and passed)
+        quote(to_bytes(parts.query, encoding), _safe_chars),
+        quote(to_bytes(parts.fragment, encoding), _safe_chars)
+    )
+
+
+def canonicalize_url(url, keep_blank_values=True, keep_fragments=False,
+                     encoding=None):
+    """Canonicalize the given url by applying the following procedures:
+
+    - sort query arguments, first by key, then by value
+    - percent encode paths ; non-ASCII characters are percent-encoded
+      using UTF-8 (RFC-3986)
+    - percent encode query arguments ; non-ASCII characters are percent-encoded
+      using passed `encoding` (UTF-8 by default)
+    - normalize all spaces (in query arguments) '+' (plus symbol)
+    - normalize percent encodings case (%2f -> %2F)
+    - remove query arguments with blank values (unless `keep_blank_values` is True)
+    - remove fragments (unless `keep_fragments` is True)
+
+    The url passed can be bytes or unicode, while the url returned is
+    always a native str (bytes in Python 2, unicode in Python 3).
+
+    For examples see the tests in tests/test_utils_url.py
+    """
+    # If supplied `encoding` is not compatible with all characters in `url`,
+    # fallback to UTF-8 as safety net.
+    # UTF-8 can handle all Unicode characters,
+    # so we should be covered regarding URL normalization,
+    # if not for proper URL expected by remote website.
+    try:
+        scheme, netloc, path, params, query, fragment = _safe_ParseResult(
+            parse_url(url), encoding=encoding)
+    except UnicodeEncodeError as e:
+        scheme, netloc, path, params, query, fragment = _safe_ParseResult(
+            parse_url(url), encoding='utf8')
+
+    # 1. decode query-string as UTF-8 (or keep raw bytes),
+    #    sort values,
+    #    and percent-encode them back
+    if six.PY2:
+        keyvals = parse_qsl(query, keep_blank_values)
+    else:
+        # Python3's urllib.parse.parse_qsl does not work as wanted
+        # for percent-encoded characters that do not match passed encoding,
+        # they get lost.
+        #
+        # e.g., 'q=b%a3' becomes [('q', 'b\ufffd')]
+        # (ie. with 'REPLACEMENT CHARACTER' (U+FFFD),
+        #      instead of \xa3 that you get with Python2's parse_qsl)
+        #
+        # what we want here is to keep raw bytes, and percent encode them
+        # so as to preserve whatever encoding what originally used.
+        #
+        # See https://tools.ietf.org/html/rfc3987#section-6.4:
+        #
+        # For example, it is possible to have a URI reference of
+        # "http://www.example.org/r%E9sum%E9.xml#r%C3%A9sum%C3%A9", where the
+        # document name is encoded in iso-8859-1 based on server settings, but
+        # where the fragment identifier is encoded in UTF-8 according to
+        # [XPointer]. The IRI corresponding to the above URI would be (in XML
+        # notation)
+        # "http://www.example.org/r%E9sum%E9.xml#résumé".
+        # Similar considerations apply to query parts.  The functionality of
+        # IRIs (namely, to be able to include non-ASCII characters) can only be
+        # used if the query part is encoded in UTF-8.
+        keyvals = parse_qsl_to_bytes(query, keep_blank_values)
+    keyvals.sort()
+    query = urlencode(keyvals)
+
+    # 2. decode percent-encoded sequences in path as UTF-8 (or keep raw bytes)
+    #    and percent-encode path again (this normalizes to upper-case %XX)
+    uqp = _unquotepath(path)
+    path = quote(uqp, _safe_chars) or '/'
+
+    fragment = '' if not keep_fragments else fragment
+
+    # every part should be safe already
+    return urlunparse((scheme, netloc.lower(), path, params, query, fragment))
+
+
+def _unquotepath(path):
+    for reserved in ('2f', '2F', '3f', '3F'):
+        path = path.replace('%' + reserved, '%25' + reserved.upper())
+
+    if six.PY2:
+        # in Python 2, '%a3' becomes '\xa3', which is what we want
+        return unquote(path)
+    else:
+        # in Python 3,
+        # standard lib's unquote() does not work for non-UTF-8
+        # percent-escaped characters, they get lost.
+        # e.g., '%a3' becomes 'REPLACEMENT CHARACTER' (U+FFFD)
+        #
+        # unquote_to_bytes() returns raw bytes instead
+        return unquote_to_bytes(path)
+
+
+def parse_url(url, encoding=None):
+    """Return urlparsed url from the given argument (which could be an already
+    parsed url)
+    """
+    if isinstance(url, ParseResult):
+        return url
+    return urlparse(to_unicode(url, encoding))
+
+
+if not six.PY2:
+    from urllib.parse import _coerce_args, unquote_to_bytes
+
+    def parse_qsl_to_bytes(qs, keep_blank_values=False):
+        """Parse a query given as a string argument.
+
+        Data are returned as a list of name, value pairs as bytes.
+
+        Arguments:
+
+        qs: percent-encoded query string to be parsed
+
+        keep_blank_values: flag indicating whether blank values in
+            percent-encoded queries should be treated as blank strings.  A
+            true value indicates that blanks should be retained as blank
+            strings.  The default false value indicates that blank values
+            are to be ignored and treated as if they were  not included.
+
+        """
+        # This code is the same as Python3's parse_qsl()
+        # (at https://hg.python.org/cpython/rev/c38ac7ab8d9a)
+        # except for the unquote(s, encoding, errors) calls replaced
+        # with unquote_to_bytes(s)
+        qs, _coerce_result = _coerce_args(qs)
+        pairs = [s2 for s1 in qs.split('&') for s2 in s1.split(';')]
+        r = []
+        for name_value in pairs:
+            if not name_value:
+                continue
+            nv = name_value.split('=', 1)
+            if len(nv) != 2:
+                # Handle case of a control-name with no equal sign
+                if keep_blank_values:
+                    nv.append('')
+                else:
+                    continue
+            if len(nv[1]) or keep_blank_values:
+                name = nv[0].replace('+', ' ')
+                name = unquote_to_bytes(name)
+                name = _coerce_result(name)
+                value = nv[1].replace('+', ' ')
+                value = unquote_to_bytes(value)
+                value = _coerce_result(value)
+                r.append((name, value))
+        return r