In current practice would there be serious jeopardy of portions of the
internet not being able to reach this address space due to bgp filters or
other restrictions? What is the smallest acceptable block of IPs that can be
announced without adverse or unpredictable results? Verio would most likely
be picking up these routes from us. I don't want to cause a religious
debate, but I am interested in what the industry consensus is.
I'm just doing some research, any comments would be appreciated.
Thanks,
Jean-Christophe Smith
Good question.
You know there are thousands of legacy /24's out there that were allocated by
IANA as /24's How can you aggregate them up if all you have is the /24?
To those who filter out /24's - how is this done - just by the netmask size?
http://info.us.bb.verio.net/routing.html#PeerFilter
That's how Verio does it, and I assume, that's how most people who filter by length do it as well.
--Phil
The longest CIDR block that all ISPs accept is a /8. Anything longer than
a /8 runs into some policy at some ISP.
There are many rules of thumb about what is acceptable to a wide range of
ISPs. Generally if you follow the number registry policies, and announce
the block delegated directly from the registry most providers will accept
it. Different address ranges have different historical CIDR lengths.
http://info.us.bb.verio.net/routing.html#PeerFilter
> That's how Verio does it, and I assume, that's how most people
> who filter by length do it as well.
We currently see 28804 /24 prefixes from our transits and peers which
are not more-specifics of another prefix that we see. (We see 127981
prefixes in total at the moment, so that's 22.5% of our table).
By comparison, we see 41066 /24 prefixes which are more-specifics for
another prefix, which is 32.1% of the table. In total, /24s account for
54.6% of the routes we see.
Of those 28804 isolated /24 routes, 946 are in "class A" space (0-127),
604 are in "class B" space (128-191), and the remaining 27254 are in
"class C" space (192-223).
The detailed breakdown by /8 is:
/8 |/24 routes | /24 routes
>with aggr. | isolated
====|===========|=============
4 39 0
12 722 0
13 1 0
15 6 0
16 1 0
17 2 0
20 5 1
24 706 117
25 0 1
32 94 0
38 50 0
40 23 0
43 1 0
44 2 0
55 1 0
57 11 0
61 281 32
62 360 82
63 1897 8
64 1693 102
65 2012 2
66 1918 250
67 328 35
68 381 13
69 206 71
80 218 197
81 233 31
82 15 4
128 68 26
129 149 3
130 60 2
131 63 13
132 11 8
134 107 5
135 6 4
136 176 2
137 88 21
138 65 1
139 41 9
140 131 1
141 125 11
142 92 1
143 46 0
144 79 7
145 32 1
146 145 124
147 54 12
148 287 4
149 68 11
150 120 1
151 58 2
152 219 2
153 57 1
154 7 0
155 125 4
156 53 5
157 39 4
158 139 1
159 127 16
160 43 3
161 56 13
162 182 180
163 59 4
164 111 14
165 129 23
166 87 9
167 285 19
168 163 12
169 86 8
170 298 17
171 8 0
192 582 4767
193 744 1703
194 672 1326
195 637 639
196 123 350
198 972 2499
199 1064 1763
200 1136 1877
201 1 0
202 1810 2229
203 1426 3775
204 1471 1488
205 1004 1028
206 1630 453
207 2116 370
208 2532 31
209 2294 485
210 761 130
211 320 60
212 642 275
213 611 360
214 15 0
215 22 0
216 2048 1203
217 629 424
218 95 16
219 33 0
220 113 2
221 13 0
222 0 1
http://info.us.bb.verio.net/routing.html#PeerFilter
That's how Verio does it, and I assume, that's how most people who filter by length do it as well.
Also worth noting that Verio does a loose-rpf check on their borders, so there's a possibility your packets will be dropped to multihomed customers who *do* have your /24 (if your best-path back to them is via verio.)..