Summary

  • IANA's IPv4 Address Space registry defines its four Status values in one footnote and leaves Designation and Date undefined, though Designation prints at least 28 distinct strings covering four different kinds of act.
  • The same publisher's IPv4 Special-Purpose registry defines all ten of its columns and forbids one boolean combination, so legibility here is chosen registry by registry rather than inherited from the organisation, the era or the format.

The premise I reserved this piece to test was wrong, and it was wrong in the publisher's favour. I expected to find the undefined columns on IANA's special-purpose address registry — the reachability booleans, the ones that read like an opinion. They are the most thoroughly defined text on either page. The undefined column sits on the older, larger, more consequential document: the registry that hands out every /8 in the IPv4 plan and tells you, in a column called Designation, what each block is for.

That page is titled "IPv4 Address Space" and notes that it was "Formerly known as IANA IPv4 Address Space Registry". Its stamp reads "Last Updated 2025-10-10" (IPv4 Address Space registry page). It prints seven columns: Prefix, Designation, Date, WHOIS, RDAP, Status [1], Note. The machine rendering prints the same seven, comma-separated, and the same first row (CSV rendering).

Row one, verbatim: 000/8 | IANA - Local Identification | 1981-09 | (empty) | (empty) | RESERVED | [2][3]. In the CSV it appears as 000/8,IANA - Local Identification,1981-09,,,RESERVED,[2][3], the WHOIS and RDAP fields empty in both renderings. Four kinds of information, then two blanks.

Exactly one footnote on that page defines a vocabulary, and it hangs off the Status header rather than the Designation header: "Indicates the status of address blocks as follows: RESERVED: designated by the IETF for specific non-global-unicast purposes as noted. LEGACY: allocated by the central Internet Registry (IR) prior to the Regional Internet Registries (RIRs). ALLOCATED: delegated entirely to specific RIR as indicated. UNALLOCATED: not yet allocated or reserved."

Nothing on that page defines Designation. Nothing states which event the Date column records. Neither of those two headers carries a footnote marker. The page answers one question about a row — what its status is — and leaves the question immediately beside it, what the block is for, to whoever is reading.

What the Designation column actually prints, in the rows this reading reached, is four different kinds of sentence in one field. Five regional-registry names: APNIC, ARIN, RIPE NCC, AFRINIC, LACNIC. Five "Administered by" compounds naming four of them. Thirteen former holder names, among them Army Information Systems Center, AT&T Bell Laboratories, Apple Computer Inc., Ford Motor Company and Daimler AG. Five purpose strings: "IANA - Local Identification", "IANA - Private Use", "IANA - Loopback", "Multicast", "Future use".

Those names are illustrations; the full list lives in the registry itself. The four counts add to at least 28 distinct strings, and 28 is a floor, not a total: the page text I reached ended before the final rows, so more may exist. No page in this family prints a row count, which means nobody can check that arithmetic except a reader doing it by hand.

The mixture is the finding, not the vocabulary's size. One kind of string records a transfer of responsibility to a regional registry. One records custody of a block assigned before the regional system existed. One is the name of a company that has nothing to do with the block now. One is a purpose that never entered the allocation pool at all. Four different acts, one column, no legend and no key.

The only text I read that touches the Date column is not in the registry at all. It is a news item dated 21 February 2008, "Update to the IPv4 registry format and content" (IANA announcement, 21 February 2008). It fixes the format — the Date 'will always be of the form "YYYY-MM"' — explains one Designation substring, and states a rule: '/8 allocations that contain legacy assignments in them that are now managed by the RIRs have the text "Administered by" in the Designation column.' It also says 'Legacy assignments refer to IP address allocations made before the RIR system was established' and 'Reserved address space held by IANA does not have a date listed.' It never says which event the date records.

That last sentence sits awkwardly against row one. The 2008 note says IANA-held reserved space carries no date; the current first row is RESERVED and carries 1981-09. My subtraction puts 17 years and 231 days between the note and the page's last update. I will not declare either statement wrong. The reserved space the note described may be a different population, or this row may have acquired a date later; both explanations are mine and neither is printed. What is printed is the sentence on one dated document and the row on another.

Now walk across the same site to the younger registry. "IPv4 Special-Purpose Address Space", Created 2009-08-19, Last Updated 2025-10-09 (IPv4 Special-Purpose Address Space). Ten columns: Address Block, Name, RFC, Allocation Date, Termination Date, Source, Destination, Forwardable, Globally Reachable, Reserved-by-Protocol. Below the table sits a Note that defines all ten, including "Allocation Date - The date upon which the special-purpose address block was allocated." It defines Forwardable as "a boolean value indicating whether a router may forward an IP datagram whose destination address is drawn from the allocated special-purpose address block between external interfaces", and Globally Reachable as the same test carried "beyond a specified administrative domain". Then it prints a rule that constrains the table: 'If the value of "Destination" is FALSE, the values of "Forwardable" and "Globally Reachable" must also be false.' The machine rendering carries the same ten headers and no prose definitions at all (CSV rendering).

Set the two pages side by side and the argument is not subtle. Same publisher, same subject matter, maintenance stamps one day apart in October 2025. One registry defines ten columns and forbids a combination. The other defines one column's vocabulary and leaves the column that says what a block is for entirely to the reader. Legibility in this family is not inherited from the organisation, the era or the file format. It is chosen, registry by registry, and the older registry — the one that governs the actual address plan — got the shorter treatment.

They do not even key the same way. The address-space page writes three-digit slash-eight labels: 000/8, 127/8. The special-purpose page writes CIDR: 0.0.0.0/8, 192.0.0.0/24. No sentence I read states a join key between them. The older page's footnote for 0.0.0.0/8 does point across — "Reserved by protocol. For authoritative registration, see [IPv4 Special-Purpose Address Space]." — but a pointer written in prose is not a key. A reader who wants to know which special-purpose rows sit inside which /8 has to expand the prefix by hand and hope the notations line up.

There is a second printed seam I am not going to close for them. The Special-Purpose Note says: "The IETF has reserved the address block of 192.0.0.0/24 for use for special purposes relating to protocol assignments. This registry contains the current assignments made by the IETF from this address block." Its own table then carries 10.0.0.0/8,Private-Use,[RFC 1918],1996-02,N/A,True,True,True,False,False — a block outside that /24, and a widely used private-use range at that. The same Note also warns that prefixes listed in the registry "are not guaranteed routability in any particular local or global context", so nobody should read those True and False cells as a statement about the live Internet. I report the sentence and the row. I do not pick which one the table means.

Here is the practical cost, and it is not theoretical. Two arguments run on this record all the time: how much IPv4 is left, and how much of it is spoken for. Neither page assists either argument the way a reader expects. Neither prints a row count, an address count or a percentage for any Status or Designation class. The only number I can offer is the arithmetic of the notation itself — one /8 is 256 × 256 × 256 = 16,777,216 addresses, my multiplication, and not a figure either page prints. The familiar 4.29 billion appears nowhere on them.

A reader who wants a reserved share must build the denominator by hand and then decide which of the four kinds of Designation string counts as "spoken for" — a judgement the page declines to make.

The duty follows the record. Whoever coordinates a unique resource and publishes the row is the party who can say what the row means, because that party writes the strings. A stranger cannot be asked to reconstruct an allocation vocabulary from four kinds of sentence and a date with no stated event, and then to build a scarcity argument on top of the reconstruction. The special-purpose registry proves this could be done, because it was done: one table over, by the same hands, on the same site, one day apart in maintenance.

Eight things I could not establish, all bounded by the documents read on 21 November 2026:

  1. No reviewed document defines the Designation column; the word appears on every row and is explained nowhere.
  2. No reviewed document states which event the Date column records. The 2008 announcement fixes a format and one exclusion, nothing more.
  3. The address-space page text ended before the final rows in this reading, so no row count, no complete Designation list and no address total is established here. The 28 distinct strings are a floor.
  4. No reviewed sentence states a join key between the three-digit Prefix of one table and the CIDR Address Block of the other.
  5. Neither registry prints a count, a percentage or an address total for any Status or Designation class.
  6. The 21 February 2008 announcement is a separate news item rather than a footnote on the registry, and no reviewed page says whether it still describes the current table.
  7. No RFC text was read for this piece; every RFC reference here is a reference IANA itself prints.
  8. No revision history or change log for either registry was read, so no re-designation of any row is established.

None of that is an accusation. No row in either table is wrong as far as this reading goes, no designation was chosen to flatter a figure, and no RFC obligation is asserted. The finding is about which page defines which column, and about what a reader holding both pages can honestly claim.

Sources: IPv4 Address Space registry page and its CSV rendering; IPv4 Special-Purpose Address Space and its CSV rendering; IANA announcement, 21 February 2008.