# When Rights Signals Disagree

A worked account of technical priority, missing instructions and the legal questions left after parsing.

By Manraj Singh Chandpuri

LEX.TXT/PROTOCOL/001

Published 2026-09-24 · Updated 2026-09-25

[Canonical article](https://manrajchandpuri.com/lex/protocols/understanding-tdm-reservation-protocol)

## Abstract

I use a controlled example to examine how the TDM Reservation Protocol resolves conflicting values and why the selected value must remain distinguishable from a legal conclusion about permission. I argue that a useful resolver should preserve the competing declarations, their sources and the rule that selected between them, while treating an unreadable carrier as a failure of observation rather than a convenient zero. The resulting record can explain the technical decision without pretending to establish authority, reconcile every external policy or determine whether the applicable copyright exception protects the proposed use.

## Permissions

I grant a worldwide, royalty-free, nonexclusive permission to crawl, index, retrieve, embed, analyse, summarise and quote my eligible original prose, and to use it for commercial and noncommercial model training, fine-tuning and evaluation, including making and retaining the copies reasonably necessary for those purposes and deploying the resulting models commercially. Retained copies must preserve the supplied author, canonical source and rights metadata, although I do not make this permission depend on a model naming me in every future answer.

My original prose in published LEX.TXT articles, including articles discussing RightSignal, as identified in the publication manifest. Embedded or linked resources are separate works and do not inherit this permission.

[Applicable permission](https://manrajchandpuri.com/lex/rights#original-prose) · [Inspect these signals](https://manrajchandpuri.com/lex/machine-signals/protocol-001)

Policy version 0.5

## Article

I want to know what happens when a publisher’s instructions disagree, because a system that reports only the winning value can make an inconsistent website appear more coherent than the evidence allows. The Text and Data Mining Reservation Protocol ("TDMRep"), developed through a World Wide Web Consortium ("W3C") Community Group, provides a useful starting point because its [May 2024 report](https://www.w3.org/community/reports/tdmrep/CG-FINAL-tdmrep-20240510/) specifies how certain conflicts are processed, while remaining a Community Group report rather than a W3C Recommendation.

I use an invented article and deliberately conflicting declarations below, so that the example tests a processing rule without implying that these declarations appeared on aninews.in or any other audited website. The distinction matters because my research finding in [NOTE 001](https://manrajchandpuri.com/lex/notes/is-a-403-a-reservation) concerns an instruction the instrument could not read, whereas this example begins with instructions that are available and inconsistent.

## The limited decision the protocol makes

TDMRep uses `tdm-reservation` to express whether text and data mining ("TDM") rights are reserved, with 1 recording a reservation and 0 recording its absence through this mechanism, while an optional `tdm-policy` points to proposed conditions. Its [processing priority](https://www.w3.org/community/reports/tdmrep/CG-FINAL-tdmrep-20240510/#sec-processing-priority) gives response headers priority over the well-known file and page metadata priority over those earlier channels, without treating a missing later property as an instruction to erase an earlier value.

Those rules give me a technical answer to a bounded question, which is what value the specified process selects from the declarations it actually receives. I still need a separate account of the proposed use, the relevant legal provision and the person responsible for the instruction before I can explain the selected value’s legal significance.

## A page that changes the answer twice

Suppose an invented resource at `example.org/news/report` is covered by a reservation in the well-known file, while its Hypertext Transfer Protocol ("HTTP") response carries a contrary value and its Hypertext Markup Language ("HTML") metadata reinstates the reservation. I would retain the successive readings in the following form, with each row identifying where the value came from rather than attributing the final result to an undifferentiated website policy.

| Encountered source | Declared value | Working value after this step |
| --- | --- | --- |
| Matching rule in `/.well-known/tdmrep.json` | `tdm-reservation` is 1 | 1 |
| Header on the article response | `tdm-reservation` is 0 | 0 |
| Metadata in the same article body | `tdm-reservation` is 1 | 1 |

My conclusion from this fixture is that the processing order selects 1, while the intermediate zero remains relevant evidence of an inconsistent deployment. If the metadata property were absent instead, the previously encountered header value would remain the working result, so a parser that replaces every missing later property with zero would be performing a different operation from the one described here.

That difference gives the example a practical use for both a publisher and a collector, because the publisher can identify the channel that unexpectedly changes the result and the collector can demonstrate which rule its implementation followed. Neither participant needs to conceal the superseded declarations in order to explain the final value, and preserving them makes a later correction intelligible.

## A file whose order changes the result

The report also assigns significance to the sequence of matching rules in the well-known file, which is why I would test an implementation with a broad rule placed before a narrower one and then reverse their order. Under its [matching instructions](https://www.w3.org/community/reports/tdmrep/CG-FINAL-tdmrep-20240510/#sec-tdm-file), the first matching rule supplies the relevant priority, so borrowing a longest-path rule from another protocol can change the result even where the same declarations are present.

The legal interest lies in the explanation a collector can give for its interpretation, because an assertion that it honoured the publisher’s most specific instruction is incomplete unless the relevant mechanism defines specificity that way. A record containing the file bytes, the matched rule and the processing version would permit the disagreement to be reproduced, whereas retaining only the final word allowed or reserved would conceal its cause.

I would therefore test the resolver against a small set of disclosed fixtures before relying on its classifications in a study, recording the expected technical value and the reason for it. That is a recommendation about inspectability, rather than a claim that these fixtures implement every part of the protocol or determine the legal sufficiency of the resulting reservation.

## A contrary policy outside the processing order

Now suppose the technical result is 0 while a separately encountered prose policy expressly objects to training, which presents a different problem from conflicting values within the specified channels. I cannot assume that the protocol’s internal order assigns a legal priority between every form of statement on the website, because the meaning, scope and authority of the external policy remain matters for the governing law.

My preferred record would retain both the technical result and the contrary statement, identify the unresolved conflict and leave any operational response visible as the collector’s choice. A blanket rule that the more restrictive statement always wins would be a proposed policy of my own, with consequences for publishers that intentionally distinguish uses, rather than something I could attribute to the processing order without support.

[ESSAY 003](https://manrajchandpuri.com/lex/essays/the-human-readable-problem) examines that allocation of interpretive responsibility in greater depth, while [ESSAY 001](https://manrajchandpuri.com/lex/essays/who-speaks-for-the-server) asks whether the person or service publishing a signal had the authority to express the rightsholder’s choice. Those questions can remain live even when there is no disagreement about what the parser selected.

## An unreadable policy leaves a different gap

An unavailable linked policy does not tell me that its conditions were empty, and the report’s [treatment of policy retrieval](https://www.w3.org/community/reports/tdmrep/CG-FINAL-tdmrep-20240510/#sec-tdm-policy) expressly preserves uncertainty about the conditions when the policy cannot be accessed or parsed. I would keep that failure beside the reservation value, because the question whether rights were reserved differs from the question whether the instrument could discover terms on which permission might be obtained.

The same discipline prevents me from turning a 403 at an instruction location into a silent declaration of zero, which was the central evidential difficulty in my aninews.in example. [NOTE 002](https://manrajchandpuri.com/lex/notes/before-the-signal-changes) develops a preservation record that allows the raw observation and later interpretations to coexist, so an improved resolver need not erase the result produced by the earlier version.

## What the selected value leaves unanswered

Article 4(3) of the [Directive on Copyright in the Digital Single Market](https://eur-lex.europa.eu/eli/dir/2019/790/oj/eng#art_4) gives an appropriately expressed reservation a role within a particular exception, while other conditions and other possible legal bases remain relevant to the use. For India, the possible application of section 52(1)(c) examined in [ESSAY 002](https://manrajchandpuri.com/lex/essays/indias-only-statutory-opt-out) requires its own statutory analysis rather than treating a European reservation mechanism as a complete account of Indian permission.

I therefore want a resolver to explain what it selected and why, while retaining the evidence that prevents its output from being mistaken for a universal licence or prohibition. The value of a processing rule is that it makes one decision reproducible, and the value of a good research record is that it shows where that decision ends.