Hacker News Reader

Comment by jsheard | Hacker News Reader

No, the issue here is that the HDMI 2.1 NDA is so strict that releasing an open source implementation is forbidden no matter how much you pay them. AMD has access to the specs, they've implemented it in hardware and in their closed source Windows driver, but they're not allowed to add it to their open source Linux driver.

Nvidia does support HDMI 2.1 on Linux since their driver is closed source (but that causes its own problems). Maybe AMD could compromise by releasing a minimal binary blob which only exposes the HDMI 2.1 implementation and nothing else.

View on HN

Nvidia's kernel driver is open source now [1], they just do the important HDMI bits in their closed source GSP firmware. Basically they moved the proprietary stuff to firmware and open sourced the rest. AMD could do something similar, but it would require a hardware change on their side (the GSP was a new bit of hardware added in Turing Nvidia GPUs).

1. https://github.com/NVIDIA/open-gpu-kernel-modules

> Basically they moved the proprietary stuff to firmware and open sourced the rest

I'm pretty sure they also moved a lot of stuff to a closed source user-space component, right?

This quote from that readme also seems to indicate a required user-space component that I'm pretty sure is not open sourced?

> Note that the kernel modules built here must be used with GSP firmware and user-space NVIDIA GPU driver components from a corresponding 590.44.01 driver release

The closed-source user-space component isn't new, the drivers always contained a kernel module and user-space libraries. Those libraries provide an OpenGL and Vulkan implementation. It's equivalent to Mesa for AMD and Intel GPUs (and the kernel driver is equivalent to amdgpu and i915 respectively).

Since it's closed we can't really know for sure if anything was moved to it from the kernel, but I think it's quite unlikely something like HDMI link setup was moved to user-space instead of to firmware.

And IIRC Intel has handled this by making their cards internally use DisplayPort then putting DisplayPort -> HDMI converters on the board.

HDMI Forum: Working hard to ensure HDMI isn't your first choice

What if a third-party reverse engineers the specifications and releases an open driver, regardless of what the HDMI Forum wishes?

yeah I am curious too. Could I legally just reverse engineer that binary and re-implement it?

The typical "clean room" process would be to have one group reverse-engineer the original and document it, then have another group of "un-tainted" people implement the spec.

This methodology has been shown to be an effective shield against copyright infringement, but it does not protect you from patent infringement. Presumably the spec is patent-encumbered specifically to prevent this type of "attack".

You also wouldn't have any rights to use any HDMI-related trademarks.

In general to avoid IP legal problems in the USA you can't do all of that yourself. Generally one party has to do all of the reverse engineering and write a specification based on that. Then another party can take that specification and write a "clean room" implementation.

https://www.allaboutcircuits.com/news/how-compaqs-clone-comp...

I've been thinking about this recently. What if one of the parties is an LLM?

Who knows, someone will have to get dragged into court to set that precedent one way or the other.

I think we’re waiting for the courts to deem LLMs able to sidestep any copyright and contract laws. If they do, artists and writers may be pissed, but engineers are gonna be lit (as long as they hate current status quo of nothing being interoperable)

Are there examples where a single person doing it gets successfully sued? It could just be that those companies were extra risk adverse so they came up with monetarily inefficient ways to defend themselves.

It's sort of the other side of that coin. There was a case where a company did it like this:

https://en.wikipedia.org/wiki/Clean-room_design

The courts said that was fine, and whenever that happens, lawyers are going to tell people to do it exactly like that since it's a known-good way to do it, whereas some other way is maybe and who wants a maybe if you have the option to lockstep the process that was previously approved?

Of course, if you do it a different way and then that gets approved, things change. But only after somebody actually goes to court over it, which generally nobody enjoys, not least because the outcome is uncertain.

Sure, but "can't" is a strong thing to say, when actually the result is thought to be legally untested.

You have to understand where "can't" comes from. It's a multi-layered system. If the government says you can do it one way and then the corporation's lawyers tell them to do it that way, corporate policy then becomes that you have to do it that way, and you therefore "can't" do it some other way.

This is why government regulations often create perverse incentives and unintended consequences. You can't just consider what the rule says, you have to consider how people are going to respond to it.

This is why e.g. the DMCA takedown process is widely abused. Do corporations have to execute obviously invalid takedown requests? Maybe not. Are most of them going to, when the consequence of doing it is harm to powerless third party individuals and the consequence of not doing it is potential liability for the corporation? Yup.

So... I ask Gemini to write a technical spec and Claude Code to implement it?

Basically a week-end project...

Everything old is new again. https://en.wikipedia.org/wiki/DeCSS

It worked out pretty okay for DVD Jon but I imagine it was a little scary for his dad and brother at the time.

I used to remember the lyrics to the 09 F9 song.

Sounds like a good job for all that AI power that is being used for BS. I wonder if we could all crowd source a driver, 100s of claude and google gemini subscriptions working towards breaking the standard and releasing 100s of different implementations that does the same.

Yeah right, 100s of Claude and Gemini subscriptions towards breaking the standard... That's how things are done. Not just one guy with a good reverse engineering skillset.

What if you crowd sourced not 100s but 1000s of Claude subscriptions. That's where the power is. You just give them a task and they just finish it for you. That's how things are done now.

Hard problem? Throw 50000s Claude subscriptions and it will kneel in front of you. Unstoppable. 50000s Claude subscriptions not enough, throw 10000000 subscriptions at it and problem solved. That's how it all works, we know this is the way to do things. Everybody knows you take a problem and throw more Claudes at it and that's it.

For example, we can do anything we want, we just need more Claude subscriptions. I couldn't do something the other day, the problem is I didn't have enough Claudes.

We just need an order of magnitude more Claude subscriptions to figure out cold fusion and unify general relativity with quantum interpretation of the world. Can you imagine what 10E10 Claude subscriptions would do with that problem? Problem stands no chance.

It is so annoying people think this is future, that this is analysis. Despicable.

What would you expect from z'ers growing up under closed magical shells doing everything for themselves (smartphone and tablet OSes) and later being utterly lost with the basics of IT.

Great, now my face hurts from laughing.

I think you misread the comment. Each person's AI agent breaks the standard once. He was not claiming they would work together. And even if he the act of translating and understanding large sums of text (binary data) seems easier to divide and concor than open ended problems like cold fusion or unifying quantum physics and general relativity.

Wow, full on delusional about how engineering work scales. Can't save everyone from themselves...

I know that HN replies must carry some substance, unlike majority of Reddit comments. But I wanted to say that this comment read line a poem to me.

I suppose you could do a clean room reimplantation, but I doubt you could advertise it as HDMI 2.1 compliant without legal repercussions.

On what basis? Trademark infringement?

Yes, that. I think you're only allowed to claim support/compliance if you're certified. And that, allegedly, means they run a couple of closed source tests and involves paperwork and NDAs.

What would the legal repercussions be against an anonymous coder who donated the code to multiple code forges? Action against the code forges themselves? I mean, not like they would be able to find the guy.

That's why you advertise it as HDMI 2.1 compatible instead. I believe there's precedence that allows that.

Yes, that might work. Strictly, HDMI is a registered trademark that might have strings but you could always say something like EIA/CEA-861... compatible instead

trademark doesn't cover descriptive language. saying it is an HDMI port is trademarked. Saying it is compatible with HDMI cables and displays is a purely descriptive statement.

it's compliant with Valve Digital Media Interface. The fact signalling is same as for 2.1 HDMI is pure accident

It most likely would prevent you from playing anything HDCP. HDCP is illegal (?) to reverse engineer, and there are special versions of HDCP2 specifically for HDMI. You need a license and a verified device for HDCP.

That might not matter much for an ordinary PC, but this Steam Machine will be competing for the living room with the PS5 and Xbox which have Netflix, Disney, HBO, etc; Not sure if things like Spotify are HDCP-protected.

It will be interesting to see how Valve works out the kinks for that. Honestly in general it'll be interesting, because putting those things on Steam Store basically turns Steam Store into a general software store instead of a game store. And the only cross-platform store at that.

With iOS and Android being broken open, you could have games be completely cross-licensed. I'd say other software too, but sadly with everything going the subscription model, you usually already have cross-licensing, in the form of an account.

How does HDCP work over DisplayPort? I guess HDCP is a different spec from HDMI itself?

Yes, HDCP is seperate from HDMI and DP.

The source and the sink need a HDCP-licence. Both devices have embbed keys that get exchanged to estabish a encrypted channel. Without the licence you can't get the required key material.

AFAIK, you can even sell HDMI devices without HDCP. Practically though, every entertainment device needs HDCP support.

it's removing HDCP protection that's problematic, not adding HDCP protection

looking at the available information on HDCP, it looks like the transmitter does not have to be authenticated - they use the receiver's pubkey, much like a web browser transmits to an HTTPS server

Part of what you're paying for is the right to use the trademarked tern HDMI, just like how the USB Consortium charges you stupid money to use the USB logo.

The suit over usage of "HDMI" in a reverse engineered version would wind up arguing whether or not HDMI is a genericised term and the HDMI Forum would lose their trademark. They will throw every cent they have into preventing such a decision and it'll get ugly

Can't you use a trademark to refer to the thing as long as it's clear you're not claiming to be them? Like if you say your PC is "IBM compatible" you're not claiming to be IBM, are you?

It's called nominative use, and describing a thing as "HDMI compatible" is permitted.

One doesn't get to use the logo or even the typeface, but that's not a dealbreaker at all for the purposes being discussed here. Words themselves are OK (and initialisms, such as "HDMI," are just a subset of words like nouns and verbs are).

The wiki has some background: https://en.wikipedia.org/wiki/Nominative_use

HDMI is patent-encumbered. The original specification has lost patent protection, but VRR and the other bits which form HDMI 2.1 and 2.2 are still protected as part of the Forum's patent pool. You could certainly try and upstream an infringing implementation into the kernel, but no one would be able to distribute it in their products without a license.

> no one would be able to distribute it in their products without a license.

In some jurisdictions, yes; however, some would probably still distribute it anyway, on purpose or not. I doubt all of them would get sued either, since lawsuits are expensive and difficult.

From my perspective, the objective is to make enforcement impractical.

> You could certainly try and upstream an infringing implementation into the kernel, but no one would be able to distribute it in their products without a license.

Isn't that actually a pretty good workaround? Hardware vendor pays for the license, implements the standard, sells the hardware. Linux kernel has a compatible implementation, relying on the first sale doctrine to use the patent license that came with the hardware, and then you could run it on any hardware that has the port (and thereby the license). What's the problem?

> relying on the first sale doctrine to use the patent license that came with the hardware

First-sale doctrine protects against copyright or trademark infringement. You might be thinking of "patent exhaustion"[1], which is a mostly US-specific court doctrine that prevents patent holders from enforcing license terms against eventual purchasers of the patented invention. There is no "transitive law of patent licensing", so-to-speak.

In this case, it would still not protect Valve if they exercise each claim in the relevant patents by including both hardware and an unlicensed implementation of the software process. It would protect end users who purchased the licensed hardware and chose to independently install drivers which are not covered by the license.

It's murky if Valve would infringe by some DeCSS-like scheme whereby they direct users to install a third-party HDMI 2.1 driver implementation on first boot, but I don't think they would risk their existing HDMI license by doing so.

1: https://en.wikipedia.org/wiki/Exhaustion_doctrine_under_U.S....

I saw chinese hw companies use "HDTV" or "HD" to avoid HDMI trademark usage.

Yep, and "HDML" on one device that would obey its user and strip HDCP from the stream when asked.

I've seen a few devices not advertising HDMI at all. Just calling it a generic "Digital Video" output.

It wouldn’t be HDMI 2.1 because it couldn’t be certified. And if you claimed it was 2.1 I imagine they would sue you.

Could it actually be made? I kind of wonder that. Like if one of the things you have to do is claim to the other device that you’re 2.1 would that get you in trouble? Or if you just advertise all the features and they each work is that good enough?

generally if something is needed for interoperability the courts only accept patents as a way to protected it (patents have a limited lifespan). However the law gets really complex and you need a lawyer for legal advice.

I think in this case you still couldn't claim it was certified. It would be on users to discover that if they plug an HDMI capable screen into that HDMI shaped port on your widget device, things just work and video shows up as expected

Note that if the protocol itself only works if the device claims certification you may be able to claim certification in the protocol. However you couldn't claim certification in marketing or any other context except where things wouldn't work if you were not certified.

They could just say "we believe we're compliant with HDMI 2.1 but are not officially certified". No lies, no claims they can't make, and nothing I can see that would introduce legal risk to folks unless there's some patent encumbered garbage in the spec.

Right. I would just advertise the features not the version number.

My only concern there is the protocol stuff I mentioned.

> Like if one of the things you have to do is claim to the other device that you’re 2.1 would that get you in trouble?

nintendo tried that with the gameboy. games had to have a copy of the nintendo logo in them. i dont think it was ever tested in court though.

Why on earth is a connector standard secret?

How else will you charge people from implementing support for it?

Well, in video land there is patent pools. For example, you pay nominal fee to download specs from iso/ice 14496-12 to learn the details about BMFF and then pay mpeg-la a couple of dollars per device of it uses an AVC / h264 decoder.

These are open standards, but mpeg-la tries to recoup some of the research costs from "freeloaders".

Open source implementations like ffmpeg are a bit of a grey area,here

That's obviously less bad, but let's not pretend this is great either.

Yes, not great indeed. This is why we have av1, ogg, etc. with most of the hard research re-done just to sidestep those pesky patents.

For now at least - for H.264 AVC, the patents are expired in most countries and most of the final US patents that may apply to AVC High profile will expire in the first half of 2026 [1].

Except in Brazil, where there are even MPEG-4 patents still in effect (expiring later in 2026) and the H.264 patents will last until the early 2030s, I think because of a rule that gave 10 years extra but is now changed but not retrospective for these patents [2].

1. https://meta.wikimedia.org/wiki/Have_the_patents_for_H.264_M...

2. https://intellectual-property-helpdesk.ec.europa.eu/news-eve...

Why not? Its not an open standard. This is the rent-seeking behavior you get under for-profit capitalist implementations. This is why we push so hard for open standards.

Uh, the HDMI forum is non-profit

Profit/non-profit isn't a big difference. Many non-profits are essentially businesses in practice (money spent/managed, the non-profit just a conduit to the for-profit companies that defacto own it), but just don't issue stock. A non-profit can act like this, and DOES. Non-profits exist in a capitalist context and inherit those norms. Again, this is why we aim for open standards.

Also a non-profit is just that, its not a charity. A charity is an entirely other classification and even those are regularly used and abused like this.

There is more than stock required to be non-profit. I suspect technically a non-profit could issue stock, though it is probably not something any would ever try.

Non-profit is a business arrangement where making money isn't the goal. There are many different versions of one though: many local clubs are a non-profit and they exist only for the benefit of their members.

Then why do they have all this?

That's meaningless, because they delegated licensing to HDMI® Licensing Administrator, Inc. And even if they are somehow a nonprofit: you are also not making any profit when all the money you retrieve via licensing fees is used to pay the royalties of the various patent holders.

Nobody cares if the mailing list where they discuss the upcoming specs is managed by a non-profit, the broader HDMI ecosystem is still a massive money grab.

It's not the connector, it's the communication protocol.

It's super lame though. It will be great to watch the downfall of HDMI Forum when their artificial dam against DisplayPort in the living room finally breaks.

Why would display port ever start taking over in the living room?

Because the manufactures don't have to pay a license fee and so once someone start using it everyone will follow and then drop hdmi. However so far nobody has cared enough to be first.

It's cheaper to implement than HDMI. So if DisplayPort ports are common on displays, devices will start using it (cheapo devices first). If DisplayPort ports are common on devices, displays won't need HDMI anymore. Plus, industry-wide, it's wildly inefficient to have one high-bandwidth video connector for monitors and a different one for TV's when the technical distinction between those is pretty much non-existent and we could scale our engineering effort across a much wider set of devices.

So, after a transition period, cost-saving will eventually lead to DisplayPort taking over.

> when the technical distinction between those is pretty much non-existent

I think CEC support is still spotty and ARC (audio return channel) isn't supported at all in DP.

Well, CEC is a huge mess and barely works[1]. You're right on ARC and eARC. I'd rather DP had a better version of both, but that wouldn't happen.

[1]: If you have a stack that works, I'm happy for you, but trust you're just lucky to have a working combination.

USB C is at least one reason that will apply constant pressure.

Well, if this were a free market, b/c there would be demand for it? I want a more standardized protocol so I need less cabling and connectors, and I want features like 4k that HDMI effectively (see TFA) does not support.

I would vote with my wallet … if I could.

Like, why do we need two connectors, for the same thing? DP is clearly technically superior.

Of course, there's a wide range of issues: there's a number of comments on this article stating how the HDMI forum is manipulating the market (e.g., by suppressing competitor connectors on the board, offering lower royalties for bugs, suppressing specifications), and then there's just getting out-competed by the litany of consumers who have no idea and do not care to know what they are buying, and marketplaces like Amazon that promote mystery-meat wares.

What is the dam against DisplayPort anyway? I never see it on TVs for whatever reason.

Actually it’s a bit odd, in my mind DisplayPort is highly associated with quality. But I don’t actually know if it is the superior connector or if it just seems that way because monitors are usually better than TVs in every metric other than size and brightness.

HDMI Forum don't like TV SOC boards that have both kinds of ports and discourage them from being made.

Also, HDMI Forum don't like converter boards that support every advanced feature at once (Variable Refresh Rate, HDR, etc.) and won't license them.

DisplayPort and HDMI kind of leapfrog each other in terms of technical superiority, so neither is definitively technically superior in the long term.

Apparently, the Hisense U8QG has DP-over-USB-C support. This might be the Trojan horse for DP in the living room.

The supported version of DisplayPort in that TV is on par (-ish) with HDMI 2.0; and not enough for HDR 4k120; which is one of the selling points of HDMI2.1.

DRM, I believe

I don't think so, DisplayPort incorporates the same HDCP encryption standard that HDMI uses.

edit: the source that I found was incorrect, and this statement is false.

DRM is optional with DisplayPort but mandatory with HDMI.

Did that change in a more recent version? According to the (admittedly old) source linked from the Wikipedia article, integrators are allowed to skip HDCP but incentivized with reduced royalties if they do support it.

https://web.archive.org/web/20081218170701/http://www.hdmi.o...

> For each end-user Licensed Product, fifteen cents (US$0.15) per unit sold.

> If the Adopter reasonably uses the HDMI logo on the product and promotional materials, then the rate drops to five cents (US$0.05) per unit sold.

> If the Adopter implements HDCP content protection as set forth in the HDMI Specification, then the royalty rate is further reduced by one cent (US$0.01) per unit sold, for a lowest rate of four cents (US$0.04) per unit.

You're right, the source that I found was incorrect.

Many TV manufacturers are part of the HDMI forum...

https://hdmiforum.org/members/

Here's a stupid question: per the site, "any entity wishing to make an active and material contribution to the development of future HDMI Specifications" can join the HDMI Forum for $15,000 p.a., and the Board of Directors is elected by majority vote by members.

Is there anything other than the money and desire to do so stopping 100 well-heeled Linux users from joining up and packing the board with open source-friendly directors who would as their first official act grant AMD permission to release its driver?

You’d want to submarine it because the forum could change its rules in “defense”.

But yes, it wouldn’t be much to do.

This sounds like what microsoft did to get their Office formats standardised by ISO. Paid membership to a bunch of folk and had the vote in favour of approving the standard. (I'm summarising *a lot*, but that's the general gist of it).

Sounds like a conflict of interest

Mass-market compatibility.

It's already difficult to find TVs with four fully-compliant HDMI ports; often you'll get a TV with one HDMI 2.1 port and three HDMI 2.0 ports, and sometimes the 2.1 port will also be the only eARC port so you have to choose between high framerates/resolutions and using a sound bar. In other words, even with just HDMI getting a decent set of ports is difficult.

The idea of TV manufacturers also adding DisplayPort ports seems ludicrous to me - not because it's a bad idea, but because I can't imagine them going to the trouble if there's no tangible demand. At best I could see them replacing HDMI ports with DP ports because there's limited space on the motherboard, but that would still require the board to have both HDMI and DP circuitry/chipsets and HDMI/DP certification/testing.

Then you have a TV with, say, two HDMI ports and two DP ports - which, for most users, means "two ports" since 99% of people don't have any hardware they want to connect to their TV that supports DP anyway.

So basically unless we start seeing game consoles, AppleTVs, and Rokus supporting DisplayPort we won't see TVs supporting DisplayPort, and we won't see any of those devices supporting DP because they don't need to - HDMI works fine for them and it's sufficiently universal.

Maybe China's new HDMI replacement will take off over there and make its way into devices over here, but I'm not holding out hope.

China's new HDMI replacement currently has no known benefit over HDMI in terms of protocol governance issues.

You're right. I should have said 'maybe China's new HDMI replacement will be managed better and take off'.

My understanding is that the HDMI 2.1 port situation on TVs is, weirdly enough, a SoC limitation from a single vendor.

Almost everyone (apart from... Samsung and LG, IIRC) is using MediaTek SoC for the brains for the TVs, and they just seem to be unable to make one that has enough bandwidth for 4xHDMI 2.1.

AFAIK LG and Samsung still handle theirs in-house (and that's why LG was the very first "big" vendor to ship 2.1 at all, and they rolled it out to all four ports even on their midrange TV's in _2019_!); and it's common to see those brands have more 2.1 ports.

This should be getting better in 2025/2026 model years, since it seems MediaTek has finally managed to ship a SoC that does it; but it's ridiculous how long it's taken.

> It's not the connector, it's the communication protocol

In particular the link training procedures needed to reliably push 48 Gbit/s over copper are probably very non-trivial, and could be considered "secret sauce".

That's done by the PHY layer, there's no need to implement that in software.

tbh it'll probably be GPMI, not DisplayPort

GPMI isn't an open standard and it doesn't support HDCP. It might end up being very popular in China but it will be a hard sell in markets that aren't primarily consuming Chinese media.

We also have a secret json schema for Dolby Vision, idk why are you surprised. This talk is about protocol, but the connector.

I wonder if the license dictates that you must use a specific language, or if they could ship that proprietary component in Javascript. My understanding is that well-written JS with a JIT runtime can be very close to native performance. Not only would that make fun of the forum's requirements, it would also provide transparency about what the proprietary module does on your system exactly

And what if they just do it anyway? What are they going to do, sue them? Make them scrub every git repository on the planet?

it will be easy to prove that it is not technically possible since Git is decentralized. but fines... oh, those fines could be enormous. possibly, AMD could get barred from implementing HDMI at all - all HDMI has to do is to stop selling the spec to AMD specifically.

Can't we just leak the spec?

Anyone can then implement opensource driver based on that and distribute it freely, since NDA won't apply to them.

Just because something is accessible publicly doesn't mean it's suddenly legal to copy it, same as it isn't OK to go into someone's house just because the door was open. Unless you're police for some weird reasons.

The problem isn’t that people don’t know how to do it.

So what, just the trademark issue for "hdmi 2.1"?

Call it a imdh driver then, nobody cares as long as it works.

No, for the resulting open drivers to not be legally dubious the spec can only be obtained by doing a clean-room reverse engineering.

Legally dubious in what sense? Leaking it might break trade secret protection, but afaik once it's public, it loses that protection, and the only one liable is the leaker. As far as I know, software per se is still not patentable even in the US since the actual source code is abstract mathematics, so it should be fine to publish the source (source code is fundamentally a detailed description of an algorithm, not a system implementing it), and there's effectively no way to stop an end-user from compiling and loading that source themselves. You could also distribute it from a more reasonable country like e.g. VLC does.

Well if you can download the source and compile it, I don't think it being legal really matters, just host it in a country that doesn't care.

Would someone doing a clean room reverse engineering be permissible to then share would they built?

Depends on the country; US, probably not. Many european countries, probably yes. Asia? Your gov will ask you why you would bother them with a stupid and meaningless question like that in the first place.

Reply on news.ycombinator.com