Pi Foundation is overcharging for MPEG2 Licenses

[RichardS]

I just went to get an MPEG2 license for my new Raspberry Pi, expecting that it would now be either FREE, or at most 0.50 USD equivelent (since the Pi Foundation has always maintained that they were merely passing on their unavoidable cost for this license, and MPEGLA has lowered this fee as of 2016)

The reason I expected that this license might also now be FREE, is that there is only ONE patent remaining in the MPEG2 patent pool, and the single remaining patent applies only to timing adjustments to TS streams - and this may not even apply to the GPU decoding methodology used in the Pi. For example, if we pass the GPU an already demuxed Program Stream (PS) then the MPEG2 HW Decoding done from that point on by the GPU does not fall under the single remaining patent at all. Dodging this single remaining patent would not be morally wrong in my view given the fact that it was issued in 1995, and kept alive way beyond it's justified 20 year term by your typical scumbag legal maneuvering to delay the issue date.

Even if the Broadcom and the Pi Foundation choose to continue to knuckle under to the patent trolls at MPEGLA, the fees have been reduced drastically.

Quoting from the new fee information on the MPEGLA web site:

'From January 1, 2016 forward, the royalty rate for MPEG-2 Decoding Products will be $0.50 per unit with right of voluntary termination on 30-days written notice, but Licensees may elect a royalty of $0.35 with right of voluntary termination on or after January 1, 2018 on 30-days written notice.'

So $0.35USD or at most $0.50USD does not justify charging folks £2.40 if the Pi Foundation is 'just passing allong the cost of the license' as they have always claimed.

Given this much lower fee, maybe the time has come to just roll the modest cost of the MPEG2 license into the overall cost of the Pi like H264 - just a thought . . .

[jamesh]

The cost of the licence isn't just the cost of MPEGLA bit, but also the cost of databasing and distribution. There are other costs as well which i cannot go into. There is also profit margin for the Rpf as well.

It's 2.40, much less than the cost of a ping of beer. Cheap.

[mahjongg]

not a PI Store topic (PI Store has been closed for a while now), so moving to general topics.

[stijn.ghesquiere]

I tend to agree with Richard.
I don't mind to pay 2.40, but it would be good if the foundation clarified that it's more or less an extra income and only a part of it is real licence costs.

The cost of the licence isn't just the cost of MPEGLA bit, but also the cost of databasing and distribution.

That't not really convincing. Yes, my grandmother would believe that, but on a technical savy forum... Running a database and doing digital distribution of such small amount of data?

There are other costs as well which i cannot go into.

This argument I like least of all. It suggest that there is someting to hide, although I you probably did mean to say that it's to complicated to explain.

There is also profit margin for the Rpf as well.

That's fine, but please emphasize that a bit more.

It's 2.40, much less than the cost of a ping of beer. Cheap.

It's not the absolute amount which is at stake, but the realitive amount charged.

Will it also mean that the foundation will keep charging for MPEG2 even if the patents are foregone?
(I can live with that, if it's well explained as an extra income for the non-profit organisation).

[stderr]

Given this much lower fee, maybe the time has come to just roll the modest cost of the MPEG2 license into the overall cost of the Pi like H264 - just a thought . . .

This link seems to have as the last known patent an expire date of Valentine's Day 2018. If that is the case, at that point there is no licence fee, then perhaps we could all just wait til then and have a celebration of freedom and then drop all charges:

https://en.wikipedia.org/wiki/List_of_U ... -2_patents

[W. H. Heydt]

I think the thread title is blown out of proportion.

That said... A lot depends on how the license agreements are written. If, for instance, agreeing to a license requires the licensee to pay for individual licenses regardless of the patent status, then the RPF is going to have to continue charging after the patents run out or stop selling licenses. It is also possible that, with the price drops happening as the patents run out that it may become economically feasible to to just include the license in the blob rather than selling them one by one. (One wonders what the license cost would be for 4+ million copies per year...to a charity.) Alternatively, once no patents are involved, so clever chap may find a way to do a FOSS hardware decoder. And if it could include the essential DECSS code, that would be a real plus.

[solar3000]

I also bought the mpeg license when I bought my first pi. But then I found out that the license was non transferable and not needed. So there's no reason to buy it. And since its not portable/ non transferable, it dies with the pi its tied to. That makes it not fair to buy.

[stderr]

I also bought the mpeg license when I bought my first pi. But then I found out that the license was non transferable and not needed. So there's no reason to buy it.

It's not needed if you don't want to use mpeg2 hardware.

And since its not portable/ non transferable, it dies with the pi its tied to. That makes it not fair to buy.

It is licensed for hardware and not to you personally.

[Electron752]

I may actually know someone how would be interesting in creating such things.

As for as dealing with the legal issues, that isn't something for me to say. As far as I'm concerned I don't know anything that has really changed and why this would be a new issue.

[Martin Frezman]

I may actually know someone how would be interesting in creating such things.

What does that mean?

[Electron752]

It just means I'm interested in codecs and such, which I honestly don't know anything much about through from an exact implementation port of view. Something like that might be a good thing to read up on.

[W. H. Heydt]

It just means I'm interested in codecs and such, which I honestly don't know anything much about through from an exact implementation port of view. Something like that might be a good thing to read up on.

The difficulty with writing a hardware accelerated codec for the Pi is that there is no publicly available documentation on the inner workings of the VC4. That means that just about the only way to write such a codec would be reverse engineering an existing one. On top of that, the instruction set for the VC4 internals isn't public either...

[Electron752]

That's why I'm not trying to read anything into anything. I just saying that something like that sounds Interesting and perhaps it would be a good area to do some research on. I always wanted to write stuff like that for the PC anyway. I even wrote a simple graphic equalizer at one point back in the dark ages of PC. It's wasn't very fast but it did kind of work.

I know a very high level of how stuff like that works, but I don't know anything about the codec specific math beyond it. Yes I know about laplace and such, but I've heard their is a lot more too it.

[Electron752]

Anyway, VC4 is closed source so I wasn't even talking about anything like that. I might even be able to get something like that to work from the ARM which is documented. During my ARM64 quest, I had VLC running in user mode without any hardware acceleration at all playing DVD quality video just for testing purposes.

[mfa298]

The cost of the licence isn't just the cost of MPEGLA bit, but also the cost of databasing and distribution.

That't not really convincing. Yes, my grandmother would believe that, but on a technical savy forum... Running a database and doing digital distribution of such small amount of data?

There are other costs as well which i cannot go into.

This argument I like least of all. It suggest that there is someting to hide, although I you probably did mean to say that it's to complicated to explain.

I would suspect a portion of the license cost goes to Broadcom for the electronics/firmware features enabled. Chances are most of the fee goes to them and they then pass on the relevant parts to patent holders etc.

Databasing and distribution will have an associated cost, it may not be that high but there is a cost involved (hardware, software, time). We have no idea what effort is involved in generating the license keys and recording them and related activities.

[W. H. Heydt]

Anyway, VC4 is closed source so I wasn't even talking about anything like that. I might even be able to get something like that to work from the ARM which is documented. During my ARM64 quest, I had VLC running in user mode without any hardware acceleration at all playing DVD quality video just for testing purposes.

That's the point. In order to run the decoding as fast as possible, it needs to be done in hardware dedicated to and designed to optimize doing it. That hardware is in the VC4. So if you want hardware acceleration on a Pi, a decoder *must* use the VC4 blocks. And that is both undocumented and very, very hard to reverse engineer.

[jamesh]

Odd thread.

Anyway, there are open source codecs that run on the ARM, on the Pi3 they are probably fast enough for 720p30, but try them and see, if you think 2.40 is too expensive. The Pi3 will run hotter so you may need a heatsink. Factor that in to costings.

Clearly the price of the licence is not just for the licence but also the manpower required to maintain the licencing system, plus some profit. Not sure why people might think otherwise.

There are aspects of the licencing system that are, for obvious reasons, company confidential (hence I don't talk about them) but incur a cost. MPEGLA get REALLY upset if you take the mick with licencing. An issue came up recently that had to be dealt with promptly with multiple engineers on the job. That cost probably $1000 in engineering time. Imagine how many licences you need to sell just to pay for that engineering time!

It's not simply "pass on the price of the licence at cost"

[W. H. Heydt]

MPEGLA get REALLY upset if you take the mick with licencing.

Not at all surprising, considering what happened with DECSS.

[jdb]

Threads like these are a bit tedious, especially when inferring that we are at fault.

The fact of the matter is, patents grant exclusivity on a specific implementation of an item of technology. When the implementation is fixed in hardware, exclusivity applies in the jurisdictions that matter (ignoring for a moment the mechanisms that FPGAs, one-time programmable CPLDs and general-purpose processors bolted to a mask ROM provide for circumvention).

RPTL has voluntarily entered into legal agreements with patent holders (read: licencing consortiums that claim to represent the several and whole interests of the numerous patent contributors) through independently negotiated contracts to offer use of the hardware decode facilities inherent in every Pi.

Long story short: We don't pay for hardware licencing costs, the end consumer does. If we did pay, each and every Pi would increase in price by approximately the (VC1 + MPEG2) licence cost. We charge a nominal fee for provision of a key that tells the firmware "yes, you have purchased a token that lets you use the MPEG-2 and/or VC-1 hardware decoders" so the firmware can use hardware decoders inherent in the SoC.

Our agreements with the various consortia are not available for scrutiny. It is what it is, deal with it (or campaign for patent reform, which I heartily encourage).