diff --git a/openerp/addons/base/ir/ir_ui_view.py b/openerp/addons/base/ir/ir_ui_view.py
index 1bf642ac202..3845d916843 100644
--- a/openerp/addons/base/ir/ir_ui_view.py
+++ b/openerp/addons/base/ir/ir_ui_view.py
@@ -18,6 +18,7 @@
 #    along with this program.  If not, see <http://www.gnu.org/licenses/>.
 #
 ##############################################################################
+import collections
 import copy
 import logging
 import os
@@ -260,6 +261,8 @@ class view(osv.osv):
             else:
                 node.set('data-oe-id', str(view_id))
                 node.set('data-oe-xpath', xpath)
+                node.set('data-oe-model', 'ir.ui.view')
+                node.set('data-oe-field', 'arch')
 
         return specs_tree
 
@@ -682,13 +685,18 @@ class view(osv.osv):
         r = self.read_combined(cr, uid, id_, fields=['arch'], context=context)
         return r['arch']
 
-    def distribute_branding(self, e, branding=None):
+    def distribute_branding(self, e, branding=None, parent_xpath='',
+                            index_map=misc.ConstantMapping(1)):
         if e.get('t-ignore') or e.tag == 'head':
             # TODO: find a better name and check if we have a string to boolean helper
             return
+
+        node_path = e.get('data-oe-xpath')
+        if node_path is None:
+            node_path = "%s/%s[%d]" % (parent_xpath, e.tag, index_map[e.tag])
         if branding and not (e.get('data-oe-model') or e.get('t-field')):
             e.attrib.update(branding)
-            e.set('data-oe-xpath', e.getroottree().getpath(e))
+            e.set('data-oe-xpath', node_path)
         if not e.get('data-oe-model'): return
 
         # if a branded element contains branded elements distribute own
@@ -702,8 +710,14 @@ class view(osv.osv):
                 if e.get(attribute))
 
             if 't-raw' not in e.attrib:
+                # TODO: collections.Counter if remove p2.6 compat
+                # running index by tag type, for XPath query generation
+                indexes = collections.defaultdict(lambda: 0)
                 for child in e.iterchildren(tag=etree.Element):
-                    self.distribute_branding(child, distributed_branding)
+                    indexes[child.tag] += 1
+                    self.distribute_branding(child, distributed_branding,
+                                             parent_xpath=node_path,
+                                             index_map=indexes)
 
     def is_node_branded(self, node):
         """ Finds out whether a node is branded or qweb-active (bears a
diff --git a/openerp/tools/misc.py b/openerp/tools/misc.py
index 65e7c38e536..bdbb4d13e0f 100644
--- a/openerp/tools/misc.py
+++ b/openerp/tools/misc.py
@@ -35,7 +35,7 @@ import sys
 import threading
 import time
 import zipfile
-from collections import defaultdict
+from collections import defaultdict, Mapping
 from datetime import datetime
 from itertools import islice, izip, groupby
 from lxml import etree
@@ -1066,4 +1066,32 @@ def stripped_sys_argv(*strip_args):
 
     return [x for i, x in enumerate(args) if not strip(args, i)]
 
+class ConstantMapping(Mapping):
+    """
+    An immutable mapping returning the provided value for every single key.
+
+    Useful for default value to methods
+    """
+    __slots__ = ['_value']
+    def __init__(self, val):
+        self._value = val
+
+    def __len__(self):
+        """
+        defaultdict updates its length for each individually requested key, is
+        that really useful?
+        """
+        return 0
+
+    def __iter__(self):
+        """
+        same as len, defaultdict udpates its iterable keyset with each key
+        requested, is there a point for this?
+        """
+        return iter([])
+
+    def __getitem__(self, item):
+        return self._value
+
+
 # vim:expandtab:smartindent:tabstop=4:softtabstop=4:shiftwidth=4: