diff --git a/database.py b/database.py
index 0a62ad1..c37369f 100644
--- a/database.py
+++ b/database.py
@@ -9,6 +9,8 @@ import time
 import logging
 import coloredlogs
 import pickle
+import numpy
+import math
 
 TLD_LIST: typing.Set[str] = set()
 
@@ -201,6 +203,33 @@ class Database(Profiler):
         Profiler.__init__(self)
         self.log = logging.getLogger('db')
         self.load()
+        self.ip4cache_shift: int = 32
+        self.ip4cache = numpy.ones(1)
+
+    def _set_ip4cache(self, path: Path, _: Match) -> None:
+        assert isinstance(path, Ip4Path)
+        self.enter_step('set_ip4cache')
+        mini = path.value >> self.ip4cache_shift
+        maxi = (path.value + 2**(32-path.prefixlen)) >> self.ip4cache_shift
+        if mini == maxi:
+            self.ip4cache[mini] = True
+        else:
+            self.ip4cache[mini:maxi] = True
+
+    def fill_ip4cache(self, max_size: int = 512*1024**2) -> None:
+        """
+        Size in bytes
+        """
+        if max_size > 2**32/8:
+            self.log.warning("Allocating more than 512 MiB of RAM for "
+                             "the Ip4 cache is not necessary.")
+        max_cache_width = int(math.log2(max(1, max_size*8)))
+        cache_width = min(2**32, max_cache_width)
+        self.ip4cache_shift = 32-cache_width
+        cache_size = 2**cache_width
+        self.ip4cache = numpy.zeros(cache_size, dtype=numpy.bool)
+        for _ in self.exec_each_ip4(self._set_ip4cache):
+            pass
 
     @staticmethod
     def populate_tld_list() -> None:
@@ -404,8 +433,9 @@ class Database(Profiler):
         pref = _par.prefixlen + 1
         dic = _dic.zero
         if dic:
-            addr0 = _par.value & (0xFFFFFFFF ^ (1 << (32-pref)))
-            assert addr0 == _par.value
+            # addr0 = _par.value & (0xFFFFFFFF ^ (1 << (32-pref)))
+            # assert addr0 == _par.value
+            addr0 = _par.value
             yield from self.exec_each_ip4(
                 callback,
                 _dic=dic,
@@ -415,6 +445,7 @@ class Database(Profiler):
         dic = _dic.one
         if dic:
             addr1 = _par.value | (1 << (32-pref))
+            # assert addr1 != _par.value
             yield from self.exec_each_ip4(
                 callback,
                 _dic=dic,
@@ -548,6 +579,9 @@ class Database(Profiler):
     def get_ip4(self, ip4_str: str) -> typing.Iterable[Path]:
         self.enter_step('get_ip4_pack')
         ip4 = self.pack_ip4address(ip4_str)
+        self.enter_step('get_ip4_cache')
+        if not self.ip4cache[ip4.value >> self.ip4cache_shift]:
+            return
         self.enter_step('get_ip4_brws')
         dic = self.ip4tree
         for i in range(31, 31-ip4.prefixlen, -1):
@@ -680,6 +714,7 @@ class Database(Profiler):
             source_match=source_match,
             dupplicate=dupplicate,
         )
+        self._set_ip4cache(ip4, dic)
 
     def set_ip4address(self,
                        ip4address_str: str,
diff --git a/feed_dns.py b/feed_dns.py
index f923831..03b9429 100755
--- a/feed_dns.py
+++ b/feed_dns.py
@@ -30,15 +30,19 @@ FUNCTION_MAP: typing.Any = {
 class Writer(multiprocessing.Process):
     def __init__(self,
                  recs_queue: multiprocessing.Queue,
-                 autosave_interval: int = 0):
+                 autosave_interval: int = 0,
+                 ip4_cache: int = 0,
+                 ):
         super(Writer, self).__init__()
         self.log = logging.getLogger(f'wr')
         self.recs_queue = recs_queue
         self.autosave_interval = autosave_interval
+        self.ip4_cache = ip4_cache
 
     def run(self) -> None:
         self.db = database.Database()
         self.db.log = logging.getLogger(f'wr')
+        self.db.fill_ip4cache(max_size=self.ip4_cache)
         if self.autosave_interval > 0:
             next_save = time.time() + self.autosave_interval
         else:
@@ -200,12 +204,15 @@ if __name__ == '__main__':
     args_parser.add_argument(
         '-a', '--autosave-interval', type=int, default=900,
         help="TODO seconds")
+    args_parser.add_argument(
+        '-4', '--ip4-cache', type=int, default=0,
+        help="TODO bytes max 512 MiB")
     args = args_parser.parse_args()
 
     recs_queue: multiprocessing.Queue = multiprocessing.Queue(
             maxsize=args.queue_size)
 
-    writer = Writer(recs_queue, autosave_interval=args.autosave_interval)
+    writer = Writer(recs_queue, autosave_interval=args.autosave_interval, ip4_cache=args.ip4_cache)
     writer.start()
 
     parser = PARSERS[args.parser](args.input, recs_queue, args.block_size)
diff --git a/feed_rules.py b/feed_rules.py
index 2b5596e..0889900 100755
--- a/feed_rules.py
+++ b/feed_rules.py
@@ -39,10 +39,14 @@ if __name__ == '__main__':
         source = database.RuleMultiPath()
 
     for rule in args.input:
-        fun(DB,
-            rule.strip(),
-            source=source,
-            updated=int(time.time()),
-            )
+        rule = rule.strip()
+        try:
+            fun(DB,
+                rule,
+                source=source,
+                updated=int(time.time()),
+                )
+        except ValueError:
+            DB.log.error(f"Could not add rule: {rule}")
 
     DB.save()
diff --git a/fetch_resources.sh b/fetch_resources.sh
index d659fbc..cb66ff7 100755
--- a/fetch_resources.sh
+++ b/fetch_resources.sh
@@ -35,7 +35,7 @@ dl http://data.iana.org/TLD/tlds-alpha-by-domain.txt temp/all_tld.temp.list
 grep -v '^#' temp/all_tld.temp.list | awk '{print tolower($0)}' > temp/all_tld.list
 
 log "Retrieving nameservers…"
-dl https://public-dns.info/nameservers.txt nameservers/public-dns.list
+dl https://public-dns.info/nameservers.txt nameservers/public-dns.cache.list
 
 log "Retrieving top subdomains…"
 dl http://s3-us-west-1.amazonaws.com/umbrella-static/top-1m.csv.zip top-1m.csv.zip
diff --git a/import_rapid7.sh b/import_rapid7.sh
index c8eacd1..993bfe7 100755
--- a/import_rapid7.sh
+++ b/import_rapid7.sh
@@ -9,7 +9,7 @@ function feed_rapid7_fdns { # dataset
     line=$(curl -s https://opendata.rapid7.com/sonar.fdns_v2/ | grep "href=\".\+-fdns_$dataset.json.gz\"")
     link="https://opendata.rapid7.com$(echo "$line" | cut -d'"' -f2)"
     log "Reading $(echo "$dataset" | awk '{print toupper($0)}') records from $link"
-    curl -L "$link" | gunzip | ./feed_dns.py rapid7
+    curl -L "$link" | gunzip
 }
 
 function feed_rapid7_rdns { # dataset
@@ -17,10 +17,10 @@ function feed_rapid7_rdns { # dataset
     line=$(curl -s https://opendata.rapid7.com/sonar.rdns_v2/ | grep "href=\".\+-rdns.json.gz\"")
     link="https://opendata.rapid7.com$(echo "$line" | cut -d'"' -f2)"
     log "Reading PTR records from $link"
-    curl -L "$link" | gunzip | ./feed_dns.py rapid7
+    curl -L "$link" | gunzip
 }
 
-feed_rapid7_rdns
-feed_rapid7_fdns a
-# feed_rapid7_fdns aaaa
-feed_rapid7_fdns cname
+feed_rapid7_rdns | ./feed_dns.py rapid7
+feed_rapid7_fdns a | ./feed_dns.py rapid7 --ip4-cache 536870912
+# feed_rapid7_fdns aaaa | ./feed_dns.py rapid7 --ip6-cache 536870912
+feed_rapid7_fdns cname | ./feed_dns.py rapid7
diff --git a/rules/first-party.list b/rules/first-party.list
index 54246cd..3092397 100644
--- a/rules/first-party.list
+++ b/rules/first-party.list
@@ -27,5 +27,5 @@ partner.intentmedia.net
 wizaly.com
 # Commanders Act
 tagcommander.com
-# Affex Marketing
+# Ingenious Technologies
 affex.org