Compulsory Licenses Discussion on Pho

Ernest Miller (of Lawmeme) and Fred von Lohmann (of EFF) have joined the fun on pho. I’m reposting their messages here – they’re really insightful.

In discussions like these, I quickly realize my limits. I can’t immediately and conclusively say anything on the technicalities they bring up – I’m not a statistics whiz, so I can’t say for sure that sampling or censuses or audits will fix everything. For instance, I don’t know if Aaron Swartz’s recent point about power laws is right or wrong or can be compensated for in some way. Hopefully, I’ll be able to do some research on it later. Perhaps Professor Fisher will incorporate some commentary on these issues into his next draft.

For now, I can just enjoy their insights and try to think them out, slowly. I’m most interested in getting the proposal on the table, so that we can start trying to figure these problems out. There is no Answer, but I think we can get a lot closer. This open dialogue is, I hope, part of that process.

I will try to continually update this. If someone would like their comments taken down, I will do so – I’ve posted everyone’s comments so far.

Update: Plenty to digest here. I have updated the post to include more of the messages. I included most every email, although I did leave out some that reiterated points already made by others, or strayed from the basic topics. If you are reading or writing on pho and think I missed something, please write to me.


Professor Lessig has a great post about a C+D recently sent to Creative Commons.  Michael Robertson is still innovating, we can still fondly remember the old – too bad it’s all gone.

Gaming Fisher’s System

As noted before, there’s been a bit more discussion about Professor Fisher’s plan. Mikael asked this question, off the privacy topic:

“If I take a copyrighted work, remove the watermark and then redistribute it (i e as unmarked data) – what happens according to Professor Fisher?”

I responded:

“Just like ballot-stuffing, I’d assume there would be penalties. The key is
that, unlike with watermarks like SDMI’s, there’s no strong motivation to
strip them out. (Note: Fisher’s proposal now uses a simple registration code, rather than a true watermark, but the practical difference is irrelevant.)”

Mary Hodder from bIPlog then wrote to pho:

“Actually, as far as motivation for stripping out watermarks or flags from
content, I would think that many different people would have incentives to strip and leave blank, or strip and replace watermarks, because they would want to help or hurt particular musicians or creators, companies, or maybe all those who are represented by the RIAA or MPAA verses those who are not.  I’m sure there are many lesser motivations that we could go on and on about.  The point is, there will be people who try and so any system will have to try to minimize this.

“The other problem with watermarks is, as far as I know, there are none
that can’t be gamed right now, and so while there are people working on
this, a system that relies on watermarks or some other way of making
a count of content flowing through routers isn’t yet feasible. And
the motivation to game the system is a primary concern, at least with the
researchers I’ve talked with about making some kind of reliable, secure watermark system for counting content.”

I then responded:


That’s a fine criticism, so let me explain what I mean a bit more.

The reason why I say there’s no strong motivation is that this sort of watermarking does not restrict use in anyway. It is most certainly a watermark in the DRM sense. So, the typical assumption that all watermarks are bad falls away.

I think the urge to game the system as you suggest is a bit weaker. It’s not that people aren’t motivated in those ways – it’s just that there’s probably less motivation to actually do something about it. DeCSS at least had some practical, tangible value to creators and users. People screwing over the RIAA/MPAA by altering the marks would just get a laugh.

At the same time, I concede that replacing the watermarks for personal financial gain is a bit more likely.  But I wonder how much one could gain financially by changing the watermarks. If one changes too few, then it is unlikely a substantial amount of people will download the song. If one changes too many, it is far more likely that s/he will be caught.

In any case, there are various ways to make it less likely, bringing it to tolerable levels.  I’m not quite sure why Professor Fisher has switched to only requiring a reg number in the actual filename. To make the system a little harder to game, a watermarking system would be superior. Yes, people could crack it – but the key is to make it hard for the average user to do so.  Even using ID3 tags or some form of metadata would be superior (and far less inexpensive than watermarks).

In addition, I wonder if a certain degree of cross-checking could be done. If the tag does not match the file, song, or artist names repeatedly, then those downloads would not be counted.

Such cross-checking would be particularly useful if the registration index were made available to digital media services.  Say someone searches for song X or clicks to download song X after searching for artist Y. The system could alert the user if the watermark does not match the song or if the search found similar songs with multiple registration codes, and then ask how s/he wishes to proceed anyway.

Of course, there are often multiple song titles and this would not work for all systems. But, perhaps it could keep misuse down to reasonable levels.