Summary

  • LACNIC 46’s public agenda says a 2025 LACNIC–MARYVA study across 18 countries or territories found a dominant regional trend of reactive online blocking, replacing a previously observed active pattern.
  • That agenda does not link the report, list the jurisdictions or disclose the cases, denominator and coding rule behind “reactive”. A versioned case matrix would let the public inspect the aggregation without revealing sensitive blocked material.

A website, a URL, an IP address and a platform are not four spellings for the same target. A domain block may remove every path under a name. An IP block may catch unrelated services sharing an address. A URL intervention can demand finer inspection. A platform restriction may operate through an entirely different legal and technical chain. Put them in one row labelled “blocking” and the most consequential part of the policy disappears.

That is why a short entry in the LACNIC 46 agenda deserves attention before the session takes place. Agenda item 9213, scheduled for 21 October 2026 from 15:15 to 16:00, lists César Díaz and Silvana Rivero for “Site Blocking in Latin America and the Caribbean”. Its English abstract says LACNIC and MARYVA conducted a study during 2025 covering 18 countries or territories. The presentation, it says, will show that reactive online blocking is now the dominant regional trend, unlike an earlier pattern of active blocking.

It also says action is typically aimed at websites, URLs, IP addresses and/or platforms in response to specific activities or events.

Those are meaningful claims. They are also all that the current public record supplies. The frozen agenda record has no linked report or presentation. It does not name the 18 jurisdictions, define “reactive”, specify an observation window within 2025, identify the source material or give a count for either class. That is a limitation of the agenda as it stands. It is not evidence that the researchers have no method, and it is not a verdict on results that have not yet been presented.

“Reactive” needs a clock and a trigger

Reactive blocking is not a visible property of a blocked identifier. It is a relationship between an event, a decision and an intervention. To code it, a reviewer needs at least a trigger date, the authority or process that converted the trigger into an instruction, the issue and implementation dates, and a rule for mixed cases. Without that sequence, a block imposed after an investigation and a standing monitoring regime can look identical at the resolver or router.

The denominator matters just as much. Eighteen jurisdictions could contain 18 legal regimes, hundreds of orders, thousands of identifiers or a smaller number of measured implementation episodes. One order may name several domains and IP addresses. One domain may resolve to several addresses. The same instruction may be applied by many networks at different times. Counting jurisdictions, decisions, target identifiers and implementing networks will therefore produce different regional pictures, even when every underlying observation is correct.

No particular denominator is inherently the right one. A study can reasonably ask how many jurisdictions use a regime, what share of observed decisions followed an event, or how many target identifiers were affected. But the answer must travel with the unit being counted. “Dominant” cannot be audited until a reader knows what is allowed to cast a vote.

The region already supplies hard cases

LACNIC’s own earlier reporting shows why a single label cannot carry the analysis. Its account of an October 2025 LACNIC 44 panel quotes a Brazilian ISP representative describing multiple court orders directed at IP addresses, domains or URLs. The same account reports that Uruguay had three blocking regimes with different origins and procedures. Participants disagreed about effectiveness and proportionality, while converging on the need for judicial oversight, time limits and transparency.

That record introduces several possible units inside one country: a regime, an order, a target and an operator implementation. It also shows why the identity of the policy setter should not be inferred from the network that carries out the measure. A court, regulator, rights holder, platform and access provider can occupy different points in the chain.

The technical object changes the externalities. A July 2024 LACNIC article distinguishes IP addresses, domains and URLs and argues that URL-level blocking can require providers to inspect packet content, with implications for privacy and end-to-end encryption. That is LACNIC’s account of the technique, not a universal claim that every URL intervention is implemented identically. Its March 2023 recommendations say blocking should not be the first option and, if unavoidable, should be temporary, limited to a domain or IP and coordinated with technical experts and affected operators.

RFC 7754 offers a useful discipline without settling the politics. It describes a blocking system through four attributes: who sets the policy, its purpose, the intended target and the Internet component used to enforce it. It then asks about scope, granularity, effectiveness and security. The framework matters here because “reactive” describes only one dimension. It does not tell a reader who acted, what was targeted, how the instruction was translated into a network control or what else the control reached.

Publish the join, not sensitive content

The missing public object is a join between the announced regional classification and the cases beneath it. LACNIC and MARYVA could provide that join through a compact, versioned coding matrix. Each row could carry:

  • a stable case identifier and jurisdiction;
  • the public legal or administrative source;
  • the deciding authority and class of implementer;
  • the trigger, issue date, effect date, review date and expiry date;
  • the target class and count, kept separate for domain, URL, IP address and platform;
  • the intervention code: active, reactive, mixed or unknown;
  • the coding-method version and observation window;
  • the source status, implementation status and any recorded collateral effect; and
  • a correction, coder-disagreement or revision history.

The matrix need not reproduce prohibited material, live target strings or confidential operator records. Stable pseudonymous case identifiers would be enough to show how rows become a total. Aggregate views should disclose separate denominators for jurisdictions, cases, targets and implementing networks rather than compressing them into one percentage.

Corrections are especially important. Orders expire, domains move, implementation may lag and later evidence can change a code. A revision should append a new state and preserve the earlier one. Overwriting the row would make the regional result impossible to reproduce at the date it was announced.

Evidence boundary

The current evidence establishes a forthcoming session, the wording of its public abstract and the absence of linked methods or case data on that agenda record. Earlier LACNIC material establishes that the region contains multiple legal origins, procedures and target types, and that technical scope affects proportionality and privacy. RFC 7754 and the Internet Society’s 2025 policy brief supply evaluation principles, including narrow scope, transparency, due process, temporary measures and attention to alternatives.

The evidence does not establish that the LACNIC–MARYVA classification is wrong, that the underlying study lacks a codebook, that all regional blocking is reactive or that any named authority or network acted unlawfully. It also does not make LACNIC the setter or implementer of the blocking decisions it studies. Those questions remain outside the public agenda and should remain outside any responsible claim made before 21 October.

Sources