A novel channel contention mechanism for improving wi-fi's reliability(arxiv.org)
arxiv.org
A novel channel contention mechanism for improving wi-fi's reliability
https://arxiv.org/abs/2410.07874
43 comments
When I worked on WiFi at Microsoft, we discovered that Apple was cheating. Apple was implemented a random backoff timer to have less range than the spec, which would mean that Apple devices win most contentions.
We decided not to copy this into Windows, since it would be a race to the bottom.
We decided not to copy this into Windows, since it would be a race to the bottom.
If it got tested and verified, wasn't their WiFi Alliance certification supposed to be revoked?
I would guess that this aspect of the spec was not well tested. It is possible to write a test that causes 100 or 1000 collisions and plots the random distribution of the backoffs, but that is pretty complicated versus just checking that 1 collision had a backoff that was within range.
What's the legality of devices intentionally violating wifi standards and causing potential issues with other devices? Is this something the FCC could act on?
There's no laws about it. It's mostly a "handshake agreement" enforced via the WiFi Alliance. You go through the WFA certification process to get the "WiFi Certified" sticker. https://www.wi-fi.org/certification
Usually large vendors try to do the right thing.
Usually large vendors try to do the right thing.
Aren't the 2.4 and 5 GHz bands among the unlicensed ISM bands?
So long as you're complying with the general rules for the ISM bands, I'd expect the FCC doesn't care about you following the details of a specific protocol.
So long as you're complying with the general rules for the ISM bands, I'd expect the FCC doesn't care about you following the details of a specific protocol.
FCC rules are they you aren’t allowed to ‘intentionally interfere’ with other users of bands.
As to what that exactly means, and where the line is drawn is up for debate. I doubt Apple would necessarily lose here, but ‘wifi jammers’ on Amazon would, for example.
As to what that exactly means, and where the line is drawn is up for debate. I doubt Apple would necessarily lose here, but ‘wifi jammers’ on Amazon would, for example.
This might be apple silently poking at the FCC et al, daring them to actualy speak up and thereby better define the rules. It's like how the marvel movies have been using the word "shit" once per movie in an attempt to force the MPAA to better articulate its rules. They dare the MPAA to slap an R rating on the next IronMan as Apple dares the WiFi alliance to disallow the wifi logo on the next iphone.
This is not in scope for the FCC. Anyone can use the ISM bands however they'd like, with certain power and bandwidth restrictions.
It's in-scope for the WiFi alliance, but I don't see this as "poking" them. It's either a bug, or it's a way to get your devices to work a bit better than the competition, but in a way that the WiFi alliance would never notice.
It's in-scope for the WiFi alliance, but I don't see this as "poking" them. It's either a bug, or it's a way to get your devices to work a bit better than the competition, but in a way that the WiFi alliance would never notice.
The question is whether abusing a standard is considered a form of interference, which is covered by the FCC. At this point though, it's just something that hasn't been tested in courts yet.
47 CFR § 15.5(b): “Operation of an intentional, unintentional, or incidental radiator is subject to the conditions that no harmful interference is caused and that interference must be accepted that may be caused by the operation of an authorized radio station, by another intentional or unintentional radiator, by industrial, scientific and medical (ISM) equipment, or by an incidental radiator.”
47 CFR § 15.5(c): “The operator of a radio frequency device shall be required to cease operating the device upon notification by a Commission representative that the device is causing harmful interference. Operation shall not resume until the condition causing the harmful interference has been corrected.”
47 CFR § 2.1: Harmful interference: “Interference which endangers the functioning of a radio navigation service or of other safety services or seriously degrades, obstructs, or repeatedly interrupts a radiocommunication service operating in accordance with this Chapter.”
47 CFR § 15.5(b): “Operation of an intentional, unintentional, or incidental radiator is subject to the conditions that no harmful interference is caused and that interference must be accepted that may be caused by the operation of an authorized radio station, by another intentional or unintentional radiator, by industrial, scientific and medical (ISM) equipment, or by an incidental radiator.”
47 CFR § 15.5(c): “The operator of a radio frequency device shall be required to cease operating the device upon notification by a Commission representative that the device is causing harmful interference. Operation shall not resume until the condition causing the harmful interference has been corrected.”
47 CFR § 2.1: Harmful interference: “Interference which endangers the functioning of a radio navigation service or of other safety services or seriously degrades, obstructs, or repeatedly interrupts a radiocommunication service operating in accordance with this Chapter.”
Your devices going faster than (but not excluding) other devices wouldn't count as "harmful interference".
If you somehow adjusted the definition so that this would count as harmful interference, then every competing service would also count (microwaves, bluetooth, baby monitors, 2.4Ghz cordless phones, even other WiFi, etc.).
If you somehow adjusted the definition so that this would count as harmful interference, then every competing service would also count (microwaves, bluetooth, baby monitors, 2.4Ghz cordless phones, even other WiFi, etc.).
You say that with confidence, but there's no legal precedent for whether that's true since it does degrade other device's performance by violating the standard. That's the question brought up that you can't answer, unless maybe you're a lawyer for the fcc.
The number of lawyers Apple can muster is an order of magnitude more than the FCC can, and the situation with Apple is worst case borderline. Apple isn’t likely to get in trouble with this unless a lot of things happen related to the situation that haven’t yet - and likely never will. Worst case, someone complains the the FCC asks Apple nicely to be more polite, and they then will be.
Intentional interference here doesn’t mean ‘slightly tweak behavior from a given standard’, it means things like ‘intentionally create nasty interference to block other users’ or ‘recklessly spam the band with enough noise no one else can use it’.
Notably, most cheap IOT wifi devices already do massive spamming of the band and terrible behavior that makes it nearly impossible for other users of the bands to work well, and they aren’t getting in trouble.
See everything from cheap Wifi security cameras to ‘smart’ fridges.
Intentional interference here doesn’t mean ‘slightly tweak behavior from a given standard’, it means things like ‘intentionally create nasty interference to block other users’ or ‘recklessly spam the band with enough noise no one else can use it’.
Notably, most cheap IOT wifi devices already do massive spamming of the band and terrible behavior that makes it nearly impossible for other users of the bands to work well, and they aren’t getting in trouble.
See everything from cheap Wifi security cameras to ‘smart’ fridges.
And this is why Apple won’t get in trouble, but someone selling a Wi-Fi jammer on Amazon will.
Was it a small difference from the spec, like an off by one error might be, or was it an order of magnitude difference that might point to it being significant and intentional? (Of course dropped 0 could also be a bug). That’s a fairly important detail for a claim like this.
Wow. Do you (or anyone) know if they still do it? Was there any pushback on them?
I don't know. It's possible that it was merely a bug, and maybe it was fixed at some point.
But it's also not something that would cause noticeable issues on a network. Apple devices performing slightly better wouldn't be a big red flag. There are many reasons that this could be the case, from a better implementation, better testing, better HW (antennas), a different OS with a better TCP stack, etc.
But it's also not something that would cause noticeable issues on a network. Apple devices performing slightly better wouldn't be a big red flag. There are many reasons that this could be the case, from a better implementation, better testing, better HW (antennas), a different OS with a better TCP stack, etc.
Why would you not race? Why be nice to Apple?
Accelerate.
Accelerate.
Because there is no race. Anyone can just set the backoff to zero and "win" by burning down the stadium so nobody else can play.
Regarding how commonly are used wifi and Bluetooth in our everyday life. I don't understand why we haven't more frequency bands available for it. It is more in the public interest than a lot of useless private initiatives like WiMAX that can easily get large frequency bands.
6GHz is definitely a step in the right direction. Also filters out all the poor people with old devices so you can enjoy radio silence on your 320MHz channel ;)
Temporary solution until within 5-10 years everyone has upgraded their routers and devices to sit on 6ghz, 5ghz, and 2.4ghz.
Eh, also doesn't go as far, which forces higher deployment densities, which helps.
The propagation differences between 5Ghz and 6Ghz are minimal compared with 2.4Ghz vs 5Ghz. In fact, given that there's a bunch of other protocol & HW improvements, it wouldn't be surprising to see identical 6Ghz and 5Ghz deployment density.
WiMAX has been dead for more than a decade now. Pretty rest of the spectrum is allocated to licensees for cellular/LTE/5G and other military applications.
> I don't understand why we haven't more frequency bands available for it
Do we really need more frequency? At this point it does not seem like the challenge for Wifi quality is actually the available frequency spectrum. In fact, I have a lot of devices still in this household that cannot connect to more than 2.4GHz and that is not a question of available spectrum but that supporting all those frequency bands apparently is too costly for some chips on the market.
Do we really need more frequency? At this point it does not seem like the challenge for Wifi quality is actually the available frequency spectrum. In fact, I have a lot of devices still in this household that cannot connect to more than 2.4GHz and that is not a question of available spectrum but that supporting all those frequency bands apparently is too costly for some chips on the market.
Thanks to DFS restrictions in the US, there are effectively only two 80 MHz channels that can reliably be used in the 5GHz band.
2.4GHz only has three non-overlapping 20MHz channels, and those can only do ~286 Mbps throughput in the best case when using 802.11ax.
The 6GHz band is finally allowing 14 non-overlapping 80 MHz channels and 7 160MHz channels, without any DFS restrictions (though some channels are lower-power/indoors only).
2.4GHz only has three non-overlapping 20MHz channels, and those can only do ~286 Mbps throughput in the best case when using 802.11ax.
The 6GHz band is finally allowing 14 non-overlapping 80 MHz channels and 7 160MHz channels, without any DFS restrictions (though some channels are lower-power/indoors only).
In any moderately dense city environment, apartment complexes, or high traffic areas you can absolutely feel the effects of a lack of frequencies.
In American suburbia it might not matter much, but it's definitely an issue for the rest of us
In American suburbia it might not matter much, but it's definitely an issue for the rest of us
From what I've read a significant contributor to that congestion is routers using too much power. A router in the middle of an 800 ft^2 one bedroom apartment in a dense apartment complex doesn't need anywhere near as much power to cover the whole apartment as does a router at one end of an 2200 ft^2 4 bedroom ranch style house in a suburb.
Out of the box most consumer WiFi routers will be configured for high power and most consumers won't even know that it is something they can change. Even if someone does know about it lowering it will probably make things worse for them unless their neighbors also lower their power.
Maybe consumer WiFi routers should also include some kind of long range low bandwidth communication method, such as LoRa, to find and communicate with surrounding routers to build a map of the routers in a general area (not necessarily a spatial map [1]) and then agree on power levels and channel assignments to avoid interfering.
[1] it would be a map of how they relate by WiFi signal strength rather than how they relate in space. So a two routers that are a couple meters apart but on opposite sides of a wall that nearly completely blocks WiFi radio frequencies would be far apart on the radio map despite being very close together spatially.
Out of the box most consumer WiFi routers will be configured for high power and most consumers won't even know that it is something they can change. Even if someone does know about it lowering it will probably make things worse for them unless their neighbors also lower their power.
Maybe consumer WiFi routers should also include some kind of long range low bandwidth communication method, such as LoRa, to find and communicate with surrounding routers to build a map of the routers in a general area (not necessarily a spatial map [1]) and then agree on power levels and channel assignments to avoid interfering.
[1] it would be a map of how they relate by WiFi signal strength rather than how they relate in space. So a two routers that are a couple meters apart but on opposite sides of a wall that nearly completely blocks WiFi radio frequencies would be far apart on the radio map despite being very close together spatially.
It really depends on the materials.
I have one bedroom apartment and the was unstable wifi in one half of it due to L-shaped reinforced concrete wall near the router. Cannot move it due to optic fiber input. To solve the issue I installed second router to have a wifi mesh.
I have one bedroom apartment and the was unstable wifi in one half of it due to L-shaped reinforced concrete wall near the router. Cannot move it due to optic fiber input. To solve the issue I installed second router to have a wifi mesh.
> Even if someone does know about it lowering it will probably make things worse for them unless their neighbors also lower their power.
Ding ding ding! The incentives are all wrong.
Plus, if more SNR is available, you can modulate more bits per unit time, then clear the air sooner.
Ding ding ding! The incentives are all wrong.
Plus, if more SNR is available, you can modulate more bits per unit time, then clear the air sooner.
Most OFDM receivers are EVM-limited though
I’m in a dense apartment block in Vienna. The walls are thick enough that I barely see two WiFi’s. Automatic band selection means we end up in different frequencies.
I'm in a dense apartment block in a sandstone tenement building in Glasgow. I can see about 30 and most of them are channel hopping permanently. The internet here is absolutely terrible over wifi, i've recently gotten a Wifi 6 router just so I could have very fast wifi in at least one room.
Not everywhere in the world uses the same concrete walls. It's quite common to see dozens of access points trying to share the same three 20MHz channels available at 2.4GHz in places like the US or dense areas in Asia.
I live in the US and I just counted 65 visible APs on 5GHz. The DFS channels aren't usable in this area, so almost all of those are trying to share the same 2 80MHz channels.
Perhaps if the bands were wider it would provide sufficient incentive for adoption, since there would be more channels available (and a wider band would presumably be easier than two separate bands that are further apart).
I think we could utilise the current bands much better than we currently do. As others have mentioned here there are a lot of devices violating the standards or not implementing the parts that would alleviate congestions/interference. I've personally seen plenty of devices that insist on 40MHz operation in an environment that can viably support only 20MHz.
Then there's also 802.11ah for unlicensed sub 1GHz networks and 802.11af for licensed networks. If implemented properly, could shift some usage to other bands (especially if those devices can't handle modern features or require excessive transmission power to be reachable).
Though giving everyone channel 13 and 14 would indeed be quite nice. It's clear that we can't get rid of 2.4GHz operation.
Then there's also 802.11ah for unlicensed sub 1GHz networks and 802.11af for licensed networks. If implemented properly, could shift some usage to other bands (especially if those devices can't handle modern features or require excessive transmission power to be reachable).
Though giving everyone channel 13 and 14 would indeed be quite nice. It's clear that we can't get rid of 2.4GHz operation.
Keep this in mind when someone tells you again what a beautiful demonstration of capitalism the "spectrum auctions" are. Most bits move over spectrum no one paid for.
On the other hand it seems like a miracle that wifi works pretty seamlessly globally, and there are no countries where you have to pay to use a proprietary service instead.
The major issue I had and I suspect this will have, is with devices that don't play by the rules. The unit economics mean the manufacturer is going to squeeze everything out, and if they can 'cheat' and claim higher numbers they will.
Back ten years ago the issue was devices ignoring CTS frames but this feels similar.