Re: TDMRep EPUB technique drafted

The biggest problem with the proposal, Laurent, is that the embedded TDM signal isn’t bound to the content in any reliable way.  Meaning, that a bad actor can simply acquire a copy of the EPUB, change the values in the metadata, and repost the newly TDM-permitted version of the EPUB for all to mine…. By doing this, the downstream miners would *not* be legally liable for having trained on the data – because the EPUB that they scanned granted them rights (and they had no way to verify otherwise).

This is one reason that using a digitally signed entity (ala C2PA) or a server-controlled entity (ala current TDM or robots.txt) are better (IMO) approaches.  On that note, we recently published 1.4 of the C2PA specification which includes supports for EPUB – see https://c2pa.org/specifications/specifications/1.4/specs/C2PA_Specification.html#_embedding_manifests_into_zip_based_formats

Since EPUB does support signatures, you could mandate that an EPUB containing TDM has to be signed.  That would resolve your problem – but would require EPUB production systems to add signature support.

Leonard

From: Laurent Le Meur <laurent@edrlab.org>
Date: Sunday, November 26, 2023 at 5:27 PM
To: public-tdmrep@w3.org <public-tdmrep@w3.org>
Subject: TDMRep EPUB technique drafted

EXTERNAL: Use caution when clicking on links or opening attachments.


Hi,

An adaptation of our TDMRep protocol to EPUB files has been requested by several participants to the CG, especially Börsenverein des Deutschen Buchhandels (Germany), Elsevier (Germany), Gallimard (France), Eden Livres (France), or Penguin Random House (USA).

I proposed a solution during the last W3C Publishing Maintenance WG, which deals with the evolutions of EPUB 3. The idea seems good to the participants.
This solution is now drafted as an evolution of our TDMRep specification.

Please read https://w3c.github.io/tdm-reservation-protocol/spec/tdmrep-epub.html#sec-epub and advise.

In this evolution of our specification, I only had to define an xml namespace, use our two favourite properties (reservation and policy), and define where they must be inserted and their scope.

Next steps:
1. Please comment on the issue I just opened https://github.com/w3c/tdm-reservation-protocol/issues/33
2. We'll plan a conference call in early December, discuss the details, and check for a consensus.
3. Once finalised, we'll replace the current final report with this new one.

Another point: those of you who participated in the webinar organised by Google received "a community feedback questionnaire about developing machine-readable means to provide web publisher choice and control for emerging AI and research use cases".

Have some of you already filled it? What was your point, if I may ask (private answers are ok)?

Best regards
Laurent Le Meur