diff --git a/django/utils/html.py b/django/utils/html.py
index de6b35dfd4..78359f97f6 100644
--- a/django/utils/html.py
+++ b/django/utils/html.py
@@ -168,7 +168,10 @@ def urlize(text, trim_url_limit=None, nofollow=False, autoescape=False):
                 url = smart_urlquote('http://%s' % middle)
             elif not ':' in middle and simple_email_re.match(middle):
                 local, domain = middle.rsplit('@', 1)
-                domain = domain.encode('idna')
+                try:
+                    domain = domain.encode('idna')
+                except UnicodeError:
+                    continue
                 url = 'mailto:%s@%s' % (local, domain)
                 nofollow_attr = ''
 
diff --git a/tests/regressiontests/defaultfilters/tests.py b/tests/regressiontests/defaultfilters/tests.py
index 94e8c43430..36f8f5851c 100644
--- a/tests/regressiontests/defaultfilters/tests.py
+++ b/tests/regressiontests/defaultfilters/tests.py
@@ -288,6 +288,10 @@ class DefaultFiltersTests(TestCase):
         self.assertEqual(urlize('usa.gov'),
             u'<a href="http://usa.gov" rel="nofollow">usa.gov</a>')
 
+        # Check urlize don't crash on invalid email with dot-starting domain - see #17592
+        self.assertEqual(urlize('email@.stream.ru'),
+            u'email@.stream.ru')
+
     def test_wordcount(self):
         self.assertEqual(wordcount(''), 0)
         self.assertEqual(wordcount(u'oneword'), 1)