TIC : changement d'échelle et nouveaux défis
On 11 December 2007 I gave a plenary lecture at Lille for Inria's fortieth anniversary. I then began writing it up as an article, kept revising it for a few years, and never finished it. Both are here: the slides as the audience saw them, and the article as it actually stands — unfinished.
It is published unfinished on purpose. It breaks off in the middle of a thought, and the page below says where. Nothing has been added to complete it.
The documents
The slides — 71 pages, exactly as built on 11 December 2007 and exactly as shown in Lille. Not rebuilt, not touched. A handout version from the same day is also here, with the progressive reveals flattened, which prints better.
The article, unfinished — 19 pages, in French, titled Changement d'échelle dans les technologies de l'information et de la communication : nouveaux défis scientifiques, éthiques et organisationnels. Four sections: Un peu d'histoire, Défis éthiques, Défis technologiques, Défis organisationnels.
Where it stops
Section 3, Défis technologiques, runs through free software, the four freedoms, the collapse of the licence economy, and the tangle of dependencies behind a routine update prompt — and then, on page 17, it stops on a bare line with no paragraph after it:
Transparence du logiciel.
That is not a conclusion. It is a note-to-self about what to write next, left in the file. Section 4 then follows and does finish, with a conclusion. So the article has an ending, but not the one it was heading for: the argument that line was pointing at was never written.
That missing paragraph is, in hindsight, where Software Heritage starts. I have been working on software transparency ever since; I just never went back and finished the sentence.
When it was written
The talk is December 2007. The article is not: I kept editing it for several years, so it is a 2007 talk carrying later revisions, and it should be read that way. The visible evidence in the PDF itself:
- page 9 dates the state of web indexing to « en 2010 »;
- a footnote on page 11 cites the European Commission press release
SPEECH/11/183— Viviane Reding on privacy by design. In the Commission's numbering the middle number is the year, so that citation is from 2011. - the LaTeX source was last modified on 2 August 2012.
So: talk December 2007; article revised into 2011, abandoned by 2012.
What was done to it in 2026
The PDF above was built in 2026, and it is not the file that came out of the 2012 build. No text was written, removed or reworded — every sentence is as it was left. What changed is that it now typesets:
- The document had not compiled cleanly since 2012. Eleven errors came from slide macros left inside the body of section 3 — the 2012 PDF exists only because LaTeX was told to carry on regardless, which is why the integral and the code listings are mangled in it. Those are repaired, so section 3 now reads as intended.
- Section 4, Défis organisationnels, was sitting after the end-of-document marker in the source. It is finished prose, about ninety lines of it, and it had therefore never appeared in any PDF at all. It has been moved back into the document. If you read the 2012 version, you have not read it.
- A cross-reference that printed "Figure ??", and the French quotation marks, now come out properly.
The work is recorded commit by commit, so the repair can be checked against the original rather than taken on trust.
The phrase
On page 10, explaining why the web still polls instead of being told what changed, the article says the current system works « assez bien », and adds a footnote because French had no phrase for the thing being described:
Good enough, comme on dit chez nos amis anglosaxons.
The same page carries the argument I think has held up best: that the waste is an externality. The cost of everyone re-downloading everything is real, but it is spread so thinly that no single actor ever feels it, so no single actor ever fixes it. That is why the better protocol lost — not because anyone rejected it.
The protocol in question is the 1998 IETF draft, and its own record is here.
This page is referenced from Good enough is not good enough, a series about why the open web is collected so badly and what it costs. The series takes its title from the footnote above.