diff --git a/dosagelib/http.py b/dosagelib/http.py
index 209a93681..98259f1db 100644
--- a/dosagelib/http.py
+++ b/dosagelib/http.py
@@ -3,9 +3,14 @@
 
 from __future__ import absolute_import, division, print_function
 
+from collections import defaultdict
+from random import uniform
+from time import time, sleep
+
 import requests
 from requests.adapters import HTTPAdapter
 from requests.packages.urllib3.util.retry import Retry
+from six.moves.urllib.parse import urlparse
 
 from .configuration import UserAgent
 
@@ -21,6 +26,13 @@ ConnectionTimeoutSecs = 60
 
 
 class Session(requests.Session):
+    """This session implements a very simple host-based throttling system: For
+    each hostname we see, we keep a record on when is the earliest time we want
+    to send the next request: If before sending a request this time isn't
+    reached, we sleep a bit until the requirements are satisfied. By default,
+    we only delay a random amount of at most 0.3sec - but some hosts might need
+    longer delays.
+    """
     def __init__(self):
         super(Session, self).__init__()
 
@@ -29,11 +41,35 @@ class Session(requests.Session):
         self.mount('https://', HTTPAdapter(max_retries=retry))
         self.headers.update({'User-Agent': UserAgent})
 
+        self.throttles = defaultdict(lambda: RandomThrottle(0.0, 0.3))
+
     def send(self, request, **kwargs):
         if 'timeout' not in kwargs:
             kwargs['timeout'] = ConnectionTimeoutSecs
+
+        hostname = urlparse(request.url).hostname
+        self.throttles[hostname].delay()
+
         return super(Session, self).send(request, **kwargs)
 
+    def add_throttle(self, hostname, th_min, th_max):
+        """Adds a new throttle for a host: Might overwrite the existing one.
+        """
+        self.throttles[hostname] = RandomThrottle(th_min, th_max)
+
+
+class RandomThrottle(object):
+    def __init__(self, th_min, th_max):
+        self.th_min = th_min
+        self.th_max = th_max
+        self.next = time()
+
+    def delay(self):
+        d = self.next - time()
+        if d > 0:
+            sleep(d)
+        self.next = time() + uniform(self.th_min, self.th_max)
+
 
 # A default session for cookie and connection sharing
 default_session = Session()
diff --git a/dosagelib/scraper.py b/dosagelib/scraper.py
index 8083f0712..99337af89 100644
--- a/dosagelib/scraper.py
+++ b/dosagelib/scraper.py
@@ -5,8 +5,6 @@
 
 from __future__ import absolute_import, division, print_function
 
-import time
-import random
 import os
 import re
 from six.moves.urllib.parse import urljoin
@@ -200,9 +198,6 @@ class Scraper(object):
                 out.warn(u"Already seen previous URL %r" % prevUrl)
                 break
             url = prevUrl
-            if url:
-                # wait up to 2 seconds for next URL
-                time.sleep(1.0 + random.random())
 
     def getPrevUrl(self, url, data):
         """Find previous URL."""