Summary

  • Four archived NANOG messages, dated 11 November 1999 to 16 April 2007, carry one argument in steadily shorter form: the body that allocates an IPv4 block is not the body that makes it usable, because usability begins with somebody else's filter change.
  • The reviewed pages attach an employer and a role to that name in exactly one document, and no reviewed page joins the NANOG correspondent to a registry employer; the only quantified figure in the whole set is an ARIN representative's remark inside a minute-taker's record.

The header prints a name; the signature prints four characters

Start with the machine part. The 11 November 1999 message prints a Date header reading 'Thu, 11 Nov 1999 07:02:27 +1000 (EST)' and a Subject line reading 'Re: ARIN to Allocate from 64.0.0.0/8'. It prints the sender as Bruce Campbell, on a personal domain that the archive reproduces in its own obfuscated notation, and it closes with '--==-- Bruce.' No employer, no organisation, no job title.

What it does carry is the distinction this briefing is about. The rendering is not continuous — the page appears to drop or reorder material near the sentence beginning 'Maybe a charge of $1/yr' — and inside that discontinuous text sit the two halves of the case. First: 'Urm, last time I checked, ARIN was not in the position to arbitarily decide which /8 (of the reserved /8s) it would start assigning from next. That decision is in the hands of IANA ne ICANN.

' Second: 'ISPs within the APNIC and RIPE ranges of 61/8 and 62/8 respectively have already gone through trying to get filters redone by US ISPs over the past few years.' The spellings are the page's. The two /8s are named; not one ISP is counted, not one network named, no duration given.

Seventeen months later, the 17 May 2001 message prints the same personal-domain sender, the Date header 'Thu, 17 May 2001 17:42:05 +1000 (EST)', the Subject 'Re: heads up on 80/8 [actually operational, so don't read this!!!]', and the signature '--==-- Bruce.' followed by 'currently not in North America, and certainly speaking for myself ;)'. The argument has moved from who decides to who is told. The page points readers at the IANA address-space list — 'this lists the various /8s and who is (nominally) in charge of 'em' — and states 'We, The Net Operators, should have this information so that we can update our filters, should we see fit to, and thus cause as little negative impact as possible.' It also asks why anyone would expect a regional registry to formally notify a forum outside its own region: 'Why would you expect the RIPE NCC to formally notify a group (nanog) that is not in its region (Europe) ?'

The 19 November 2003 message prints 'Wed, 19 Nov 2003 12:36:04 +0100 (CET)', 'Re: New IANA IPv4 Allocations', a variant personal-domain sender and again '--==-- Bruce.' The subject has changed; the position has not. It records that, as far as this writer has watched, the allocating body 'has made a public post about new allocations being made to the RIRs to NANOG (and other forums) within a reasonable timeframe' — and then concedes the race to a third party's 'every half hour page scraper', which beats 'the authoritative people to the punch' because it does not face 'the same unseen formalities that the IANA/RIRs need to go through'.

The 16 April 2007 page is the one that stops looking like a message at all. No From, Date or Subject block; a thread label, 'RE: Question on 7.0.0.0/8', and the byline 'Bruce Campbell 16 Apr 2007'. No address for him anywhere on the page. Its body contains the page's own ellipses and an apparent rendering break, so it is not a continuous message either. In it: "IANA's role in this should be 'Ugh. Here Big Block. Go Talk to RIR.' imo of course ;)" and, for anyone wanting an authoritative allocation directory, 'Been There, Done that. See each RIR's FTP site, updated daily. There's even a standardised format between the RIRs, unlike the whois output.' Signature: '--==-- Bruce.'

Four documents, four dates, one signature, and zero organisations printed on any of them.

One document prints a job, and it is not about addresses

The 9 February 2000 apnic-talk post is the exception that frames the whole record. It prints 'Wed, 9 Feb 2000 12:40:17 +1000 (EST)', the Subject 'Re: [apnic-talk] a proposal agendum for policy SIG at Korea', and a sender address on the registry's own domain. Its signature block prints, line by line: the name; a contact telephone number, withheld here; 'Systems Administrator'; 'Regional Internet Registry'; 'Asia Pacific Network Information Centre'; 'For the Asia Pacific Region'.

That is the only reviewed document in which the name attaches an organisation and a role to itself in the author's own signature. Its body is about whois database mirroring, and it asks participants to return the thread to its stated topic, 'Handling of Personal Information in the APNIC database'. It contains no allocation claim, no exhaustion claim, nothing about filters. The one page that says who this person worked for says nothing about what the argument says.

The APNIC 11 DNS SIG minutes add a second, weaker link and are worth reading for what kind of document they are. Headed Tuesday 27 February 2001 at the Putra World Trade Centre in Kuala Lumpur and closed 'Minuted by: Gerard Ross', they credit 'Bruce Campbell, APNIC' on three presentations. The third, 'Legacy transfers', is described as having 'outlined a possible RIR approach to dealing with legacy "swamp" address space by transferring registration data to the regional databases from the ARIN whois'. That is a minute-taker's third-person summary of a presentation. It is not the correspondent's prose, and it should never be quoted as if it were.

The only number in the set belongs to somebody else

The minutes contain the single quantity anywhere in this record: 'An ARIN representative noted that there are approximately 49 /8s affected.' Read the sentence's owner. Not the presenter. An unnamed representative of another registry, speaking in a question period, recorded by a minute-taker, about legacy registration data. The same passage records that the re-delegation work 'has been negotiating with vendors for an appropriate server platform ... although there is currently no definite timeline', and that the IETF 'still needs to resolve key handling issues'.

So the arithmetic of usability is missing everywhere it would matter. Across all reviewed pages there is no count of networks affected by a filter change, no duration for such a change, no notification window, no due date, no monitoring baseline. The claim that a block becomes usable downstream is made four times in eight years and never once quantified by the person making it. The only 49 in the file is borrowed from a voice at the second remove.

What the files do not carry

Here is where the label and the record part company. The RIPE NCC ncc-services-wg archive index for October 2003, as inspected, contains no message from him at all. No reviewed page prints a RIPE NCC affiliation line for this correspondent. The APNIC role line is real, self-attested, and dated 9 February 2000; a RIPE NCC role line is not in the record.

Three documents bear directly on that gap and were seen only as search-result text, not opened as pages: a 2002 mailing-list post carried from an address on the registry's ripe.net domain whose signature names RIPE NCC and NANOG 25; the RIPE 43 attendee list printing 'Bruce Campbell' with 'RIPE NCC'; and the APNIC 1998 annual report listing 'Bruce Campbell, Systems Administrator'. Any of the three, read whole, could close the chain between the NANOG correspondent and a registry employer. None of them was read whole for this briefing, so the chain stays open here, and I will not pretend a surname plus a domain is a personnel record.

The ledger and the filter

What the record does establish is a boundary, stated four times, in four registers. Allocation is a ledger act: an administrative decision inside a registry's paperwork, defended in 1999 against a claim that a registry was overstepping, and reduced by 2007 to a shrug — here is a big block, go talk to your registry. Usability is a filter act: a change made by an operator who has no obligation to read the allocation announcement at all, which is why the writer wanted the allocation list in operators' hands and not only in registries' archives.

Our reading of that boundary is ours, not his: a record that logs allocations but not the operators who make blocks effective has recorded authority at the wrong layer. It is the sort of error a ledger makes naturally, because a ledger counts what it hands out and cannot see a filter rule. The 2003 message says the quiet part out loud — the third-party scraper beats 'the authoritative people to the punch' — and in doing so separates being first from being authoritative, which are two different claims that institutions often let merge.

My own arithmetic, labelled as mine and held apart from every printed figure: the reviewed set is seven pages; printed affiliations resolve to one organisation, APNIC, in two documents, and to zero organisations across the four NANOG messages; the first and last printed dates are 11 November 1999 and 16 April 2007, which I subtract as 7 years, 5 months and 5 days; the printed time offsets are +1000, +1000, +0100 and none, and no reviewed page explains the change.

What closes it

The archive did its job on the argument and stopped at the door of the person. Anyone holding these seven pages can learn precisely where a registry's authority ends and an operator's work begins, in the words of somebody who had to live with both. The same reader cannot learn who was speaking, and cannot honestly infer it: the four NANOG pages print a signature and no employer, the one page with an employer is about database mirroring, and the minutes speak about him in the third person. What the record needs is not more interpretation but three specific documents, opened whole.

Until then the finding stands as it is — an eight-year case, made in public, under a signature four characters long.