Summary

  • LACNIC's technical reports print a fixed four-column credit block, and Guillermo Cicileo's name appears in six different columns of it between 2017 and 2024; no reviewed document defines what the coordination-and-revision column requires of him.
  • The regional RPKI figures LACNIC presents as its published work are bylined Erika Vega, while the coordination credit is printed only in a separate post dated 10 February 2026 — one citation whose two halves live on two pages five months apart.

The instrument is four lines long

Start with the least glamorous page in Latin American routing research: the front matter of LACNIC's IPv6 CPE report. It prints, in the publisher's own words, "Author: Alejandro D'Egidio. Coordination and Revision: Guillermo Cicileo. Edition: Communications Area. Department: Technology. 2022" (IPv6 CPE Report, provider-reported excerpt from search-indexed text).

That is an instrument, and a good one. It answers three questions a stranger actually needs: whose sentences these are, which department owns the output, and when the edition closed. It raises a fourth question and walks away from it: what did the person in the coordination slot do? Nothing on that page says. Nothing in the version table says either, and the version table is itself a small lesson in how records behave — it prints Version 7 dated 09/09/2021 under a 2022 imprint, and the reviewed page does not explain the relation between the two dates.

Now put the same instrument beside itself. In the Connectivity 2022 study the coordination column is not a person at all: the front matter reads "Autor: Agustín Formoso y Guillermo Cicileo. Coordinación/revisión: Área de Comunicaciones. Edición: Septiembre 2023. Área: Área de Tecnología" (Estudio de Conectividad 2022). In the IPv6 deployment forecast the column is a person again, and a fifth line appears: "Coordinación/revisión: Guillermo Cicileo. Edición: Área de Comunicaciones. Supervisión general: Área de Tecnología" (Evolución regional y futuro a corto plazo del despliegue del IPv6).

So the block is not a table of duties. It is a set of columns, and the columns change occupants and sometimes count. On one document coordination belongs to a communications department; on another it belongs to the head of the research area while general supervision belongs to the technology area; on a third the reviewer is a named person and the author is an outside academic. Three documents, one publisher, three different meanings for the same slot — and no page in the set states which meaning governs.

Six appearances, six different obligations

The inventory is worth laying out because it is the whole argument in miniature. Six slots, six documents, six different things a reader is implicitly asked to trust.

On 31 August 2017 the LACNIC Labs blog ran a post under a byline that prints name, role and handle together: "Guillermo Cicileo — Líder de Investigación y Desarrollo en infraestructura de Internet — @ssrlac" (Medidas paliativas ante el agotamiento de IPv4 en LACNIC). That is the tightest instrument in the set: one person, one title, one date, one argument. And the argument is an instruction, not a credit. It tells internet service providers not to deploy CGN, LSN or NAT444 without putting IPv6 in front of end users at the same time, on the reasoning that more than half of user traffic would by then run native IPv6 and bypass the carrier-grade NAT equipment entirely. It points operators at 464XLAT (RFC 6877) and SIIT-DC (RFC 7755 and 7756). This is the one place in the whole corpus where a person's own sentence carries a directive that an operator could implement or refuse.

In the Connectivity 2022 study, edited September 2023, Cicileo is an author beside Agustín Formoso, and the research measures something measurable: traceroutes toward addresses in each country, roughly 5,500 networks, with a paths view added over the 2020 campaign. The study reports that countries with IXPs or large operators keep more than 90% of traffic inside the country with latencies typically under 40 ms, and that IXPs typically route more than 40% of measured traffic (source). Here the credit is authorship and the reviewer is a department.

On the CPE report he is the reviewer and the author is someone else, and the survey underneath is worth quoting because it shows what a review is being asked to cover: 75 ISPs identified, 36% of them providing IPv6 service, DS-Lite supported by 89.7% of identified CPEs although only one ISP reported using it, Huawei accounting for 15 of 60 identified CPE models — 25% — and present in 32.7% of ISPs, with ZTE at 11.5% (source). Those are not decorative numbers. A vendor-concentration figure with a single-ISP usage anomaly beside it is exactly the sort of claim a reviewer is supposed to interrogate, and the reviewed page does not record whether anyone did.

On the IPv6 forecast the arrangement is the most interesting of the six, because the author is outside the organisation — Emiro A. Molina-Cuevas, Ph.D. — and the reviewer is the head of research. The method is stated: monthly estimates derived from Google IPv6 statistics published through LACNIC's open data. The projections are dated and therefore falsifiable. South America reaches 51% of users around late January 2027 (51.07%, error 4.74%); the Caribbean around October 2028 (51.31%, error 20.26%); Central America 51% around December 2023 (51.45%, RMSE 6.54%) (source).

On 7 December 2022 the LACNIC Labs blog co-bylined a RIPE Atlas deployment report for Colombia, printing "Elisa Peirano — Asistente de I+D" above "Guillermo Cicileo — Líder de Investigación y Desarrollo en infraestructura de Internet". The report hands out hardware and then counts what survives: nine probes distributed at LACNIC 37 in Cali, and, as of 10 November 2022, 12 connected, 4 disconnected and 32 abandoned in Colombia, with several disconnections attributed to USB stick and file-system failures (source). A co-byline on a document that prints its own attrition is a different kind of credit from a co-byline on a headline percentage.

And on 27 June 2024 a mailing-list message carries a self-signed block: "Líder de Investigación y Desarrollo en infraestructura de Internet / Head of Internet Infrastructure Research and Development / LACNIC" in two languages under one name (nogcolombia archive). No intermediary, no editor, no front matter. Just a person printing his own titles on a machine-dated message.

Two titles and one page that cannot date itself

What holds these documents together is a printed name and a stated affiliation, and the strength of that link differs by document. It is strongest in the 2024 signature block, because he wrote it. It is strong in the 2017 post and the 2022 Labs report, which print name, role and handle together. It is weakest on the staff biography that describes him as "currently Head of Internet Infrastructure Research and Development at LACNIC", a member of the LACNOG board and coordinator of the Training Task Force (staff biographies) — because that page prints no date at all, so it cannot anchor the title to any year, and it cannot be used to date anything else.

The title record also moves. An ITU-T workshop biography, on a page belonging to a workshop dated 30 May 2018, lists him as "Head of Security, Stability and Resiliency, LACNIC" and records a seat on the FLIP6 evaluation committee from 2007 to 2017 (ITU-T biographies). Between that and the later Spanish and English renderings of the research role, the record shows a change of title and says nothing about when or why. That is not a scandal; it is a gap, and it matters here for one reason: a credit block is only as strong as the reader's ability to know which office the credited person held on the day the edition closed.

The citation that lives on two pages

Here is the sharpest version of the problem, and it concerns the numbers this region actually argues about.

On 17 September 2025 LACNIC's blog published an analysis of RPKI adoption, and the byline is "Por Erika Vega". The figures are precise and the denominator is stated: routes classified valid, not-found or invalid, not allocated or assigned address space. Over a 2019–2025 window, globally valid IPv4 routes rise from 14.39% to 51.14% and IPv6 from 20.19% to 57.01%, while not-found IPv4 falls from 85.13% to 42.21% and IPv6 from 79.04% to 36.22%. For Latin America and the Caribbean the same post prints IPv4 valid routes rising from 9.89% to 46.10% and IPv6 from 16.56% to 60.87% (Evolución global y regional de la adopción de RPKI).

That page is, on the measurement axis, more careful than most of what circulates in routing-security commentary. It says what the denominator is. It says what the window is. It does not say who coordinated the study.

For that, a reader has to find a different post, published 10 February 2026 under the bylines of Carlos Martínez and Gerardo Rada, which states that the work was "elaborado por Erika Vega, chair del Grupo de Trabajo de Enrutamiento de LACNOG, con coordinación de Guillermo Cicileo (LACNIC)" (RPKI: herramienta clave para la seguridad del enrutamiento celebra 15 años en la región).

Two halves of one citation, five months apart, on two pages, neither of which points at the other in the reviewed record. My own arithmetic, not the publisher's: 17 September 2025 to 10 February 2026 is roughly five months. The consequence is not that anyone hid anything. The consequence is structural. A reader holding the measurement page cannot learn from it who coordinated the work, and a reader holding the February page learns who did without learning a single number. If coordination means something checkable, the page that depends on it should carry it.

The forecast that can be graded by the calendar

Of the six documents, the one that best tests what a review slot is worth is the forecast, precisely because it invited the future to disagree with it.

Central America was projected to pass 51% of users around December 2023, with a printed RMSE of 6.54%. That date has passed. The reviewed set contains no retrospective scoring of the projection, no note recording whether the estimate landed inside its own error band. South America's late-January-2027 date and the Caribbean's October-2028 date are still open.

I am not alleging that nobody checked. I am stating what the record shows: a document whose author sat outside the organisation and whose reviewer sat inside it made a falsifiable promise about a specific month, the month arrived, and no reviewed page records what the promise was worth. A forecast is the one credit-block artefact that can be graded without asking anyone's opinion, and it is the one artefact where the review slot would have a visible output if it had one.

When the measurement measures its own losses

There is a counter-example in the set, and it deserves its place. The Colombia RIPE Atlas report counts what failed: 32 abandoned probes against 12 connected, as of 10 November 2022, with USB stick and file-system failures named as causes for several disconnections (source).

That is a small thing and it is the right instinct. A measurement programme that prints its attrition is telling a reader where the sample went thin. It is also the only reviewed document in which the credited people are reporting a deficit rather than a result. If LACNIC wants a model for what a credit block should signal, it is not the headline percentage; it is the abandoned-probe count.

What the block can and cannot attest

The honest boundary has to be stated plainly. An absence of a printed duty across ten documents is a fact about ten documents. It is not proof that no duty exists inside LACNIC. No reviewed page in this set prints a review checklist, a review log, a sign-off record, or an editorial-policy statement for the coordination and revision column, and none of the underlying PDFs was read as a page image — figure captions, chart values and footnotes were not inspected. The block's vocabulary is real, its dates are printed, its names are printed, and its obligations are not.

What a credit block can do is narrow and valuable: it can tell you who to ask. What it cannot do is tell you what was asked of them. For a registry whose research output becomes other people's evidence, that difference is the whole ballgame. The names in the block are the least interesting part of it. The interesting part is that the region's routing-security numbers have a coordinator who is off-page, a reviewer whose duty is undefined, and a forecast whose date has already arrived.

If Cicileo's coordination is worth printing, it is worth defining. Four lines of front matter already exist. A fifth line saying what coordination covers would cost a paragraph and would make every number in that corpus easier to stand behind.