codec development involves real research work [LWN.net]

codec development involves real research work

Posted Feb 1, 2018 1:31 UTC (Thu) by tialaramex (subscriber, #21167)
In reply to: codec development involves real research work by jensend
Parent article: Chiariglione: A crisis, the causes and a solution

"You don't get a Fields Medal or a science Nobel by puttering around by yourself for an hour after your day job"

Mmm. So, the most famous example of someone who _didn't_ get the medal is Andrew Wiles. Now, in a sense Wiles looks like an example of the phenomenon you're talking about, working at prestigious universities such as Oxford and Princeton. So this was his day job, right?

Except, nobody was paying Wiles to do breakthrough work on Fermat's Theorem, it was a crank hobby, people with no serious mathematical background tinkered with it, you couldn't get serious research money to keep chipping away at this century's old puzzle, even if it seemed tantalisingly close. Wiles had to _pretend_ to be continuing his earlier work in related fields, lie to academics he worked alongside, parcel out small discoveries he'd actually made years ago, rewrite papers, and so on, looking like a _failure_, somebody who was destined to dead-end to cover up the fact that all he was really spending his time thinking about was Fermat and how to prove something called the modularity theorem which was in Wiles' area and if it's true (which it is) proves Fermat was right.

I'm very familiar with the "good cause" argument for why bad ideas are actually OK, there's a play called "Peter Pan" you've probably seen adapted in some form, and in my country (only) a famous children's hospital has been granted a perpetual royalty right in this play, transcending any ordinary "intellectual property" protection such as copyright licensing, on the argument that it's a good cause, so we should keep the money flowing forever. I do not need to think that treating sick children is a bad idea to recognise that this decision was inappropriate. Likewise then, patent licenses aren't free money, somebody is paying those license fees, if we need lots of graduate students let us instead be honest and use tax revenue to pay them, not try to somehow shackle the research work to codec patent licenses, as if these have anything in common.

to post comments

codec development involves real research work

Posted Feb 1, 2018 17:23 UTC (Thu) by jensend (subscriber, #1385) [Link]

Wiles was working with experts in the field on elliptic curves and Iwasawa theory as a (paid) graduate research assistant by 1975. After he finished his PhD in 1980, he was at four different academic institutions (IAS, Princeton, IHÉS, ENS), working full time on elliptic curves with the world's best and being paid to do so, before he heard in 86 the epsilon conjecture had been proved and decided to focus on the modularity theorem (which would also prove FLT).

"Cranks" weren't working on the modularity theorem. This was not a crank thing to be doing. It was precisely the kind of mathematics Princeton (and later Oxford) were paying him to research, discuss with colleagues, and lecture on. The secrecy for seven years wasn't because this was a strange hobby. It's because 1) he didn't want to let hints slip to others and get beat to the punch, he wanted full credit and 2) he knew that the normal thing would be to focus on producing a steady stream of smaller results which would erode the seemingly insoluble problems over time rather than bang your head against something when there's not visible publishable progress. After he publicized his work, it took another two years of working with colleagues who helped him correct errors and finish the proof.

Yes, he wasn't going around telling people "I'm focused on Fermat's Last Theorem" for those seven years. But he was doing exactly what he was paid to do full time, which he'd been paid to do full time for twenty years by the time the corrected proof was finished, and he was able to do it because he'd been mentored by and worked with the best minds in the field.

Your Pan diatribe is a total irrelevancy.

Even though grants have to be part of the picture, that has its own tradeoffs. Saying "divorce the rewards of good research from the researchers and instead use tax money, compelling people via the threat of imprisonment to pay for lots of researchers, without regard to whether the researchers achieve results or whether those forced to pay actually use the results themselves" is not a flawless answer to the problem at hand.

And though I think patents and royalties aren't inherently evil and that their merits can exceed their ills in some fields, I agree that using them for the basic formats of information exchange creates far too many problems. (Even if MPEG had a saner IP policy and even if the licensing had been administered more rationally and fairly than MPEG LA &c have done it wouldn't have been enough.)

I've been following Xiph and their efforts closely for fifteen years. I'm personally very glad that with Opus and AV1 and with the massive HEVC defections it looks like open and RF formats will finally win out. I'm just saying we have to recognize there are real tradeoffs here rather than claim innovations will henceforth arise out of vacuum. Xiph folks would be the first to say their efforts have relied on others' academic work (e.g. Fisher, "A Pyramid Vector Quantizer," published in IEEE Inf Thy in 85, is vital to Opus), and Xiph's formats were a third place niche until they had plenty of corporate support.