The same packets, a very different channel
Everything you have learned so far still holds in the air. A packet is still a postcard with an address; layering is still nested envelopes; a frame still carries its addresses up front. When your laptop talks over Wi-Fi or your phone over a cell tower, the IP packet inside is exactly the one a wire would carry. What changes is only the bottom of the stack — the physical and link layers, the part that actually pushes bits across the gap. So this whole rung is not a new Internet; it is the same Internet with a radically harder last hop.
On a wire, the channel is almost boring in the best way. A copper pair or an optical fiber is a private, shielded tunnel: the signal you send is the signal that arrives, only weaker, and almost nothing else gets in. The bit error rate on a good wired link is so tiny — often fewer than one bad bit in a billion or a trillion — that higher layers can mostly pretend errors never happen. A wireless link throws that comfort away. The air is shared with every other radio, every microwave oven, every wall, and the signal does not travel in a tidy line. The single most important sentence in this rung is this: wireless is not a worse wire, it is a fundamentally different kind of channel.
Keep three words separate as you climb, exactly as you learned earlier. Bandwidth is how wide the pipe is (bits per second), throughput is what you actually get after losses and retries, and latency is how long one bit takes to arrive. On wireless, throughput swings wildly second to second as conditions change — yet none of that lets you beat the speed of light, so more bandwidth still does not cut latency. Holding those three apart will keep you honest about what "faster Wi-Fi" or "5G" really promises.
Signals fade, and noise never sleeps
Start with the gentlest problem: the signal simply gets weaker with distance. This is attenuation, and on radio it is brutal. Power does not drop linearly with distance but roughly with the square of it in open space — double the distance and the signal can fall to a quarter, and walls, floors, and your own body steal even more. That is why one room away from the router you have full bars, and two concrete walls away you are crawling. A wire attenuates too, but predictably and slowly; the air's attenuation changes the moment someone closes a door or walks between you and the access point.
A faint signal would still be fine if it were the only thing in the air. It is not. The receiver also hears noise — thermal hiss in its own electronics — plus interference from every other device on the same frequencies. What actually decides whether a bit survives is not the signal's raw strength but the signal-to-noise ratio (SNR): how loud your signal is compared to everything else. A clear voice in a quiet room is easy to hear; the same voice at the same volume in a loud bar is lost. Shannon's old result from the foundations rung bites hard here: the capacity of a channel rises with SNR, so as your signal fades and the noise stays, the honest maximum data rate falls. That is why a weak Wi-Fi connection does not just feel slow — the link literally negotiates a lower speed to keep the bits readable.
Multipath: the signal arrives more than once
Now the strangest wireless problem of all, and one that has no equivalent on a wire. Radio waves bounce. The signal that leaves your phone does not take one path to the tower — it takes many. One copy flies straight; another reflects off a building; another bounces off the floor and the ceiling. Each reflection travels a slightly different distance, so the copies arrive at slightly different times. This is multipath propagation: the receiver hears not one signal but a blurry pile of echoes of the same signal, smeared across time.
Think of shouting one word down a canyon: you hear the word, then its echoes, and if the echoes overlap the next word you shout, the listener hears mush. Multipath does both bad things at once. When two copies of the same bit arrive almost lined up but slightly out of phase, they can cancel each other — a dead spot where moving your hand a few centimeters restores the signal (this is fading). And when the echo of one bit smears into the arrival of the next bit, the symbols blur together, which is called inter-symbol interference. The wire never has this problem because there is only one path down the tunnel.
WIRE: sender ----------------------- receiver
one path, one clean copy arrives
AIR: sender ===> direct path ==========> receiver
sender ===> bounce off wall =====> receiver (a bit later)
sender ===> bounce off floor ====> receiver (later still)
receiver hears: | | | . | | . . | same bits, smeared in timeHere is the beautiful twist: modern radios turned this curse into a feature. Wi-Fi and 4G/5G use a clever scheme (OFDM) that slices the channel into many narrow sub-channels, each slow enough that the echoes settle before the next symbol — so smearing stops hurting. Better still, with several antennas (MIMO), a receiver can hear the different reflected copies as separate streams and add them up, or even send several independent streams at once. The echoes that once destroyed the signal become extra capacity. You do not need the math here; just hold the idea that multipath is the defining hard problem of the air, and that beating it is most of why wireless got fast.
The collision you cannot hear
Now recall the shared-wire Ethernet you met in the LAN rung. When many machines shared one cable, the rule was CSMA/CD: listen before you talk, and if you detect a collision while sending, stop and back off. The key word is detect — on a wire a sender can listen to the cable at the same instant it transmits and notice two signals smashing together. On the air, that trick is impossible. A radio transmitting its own signal is shouting at full volume; the faint, distant signal of another sender's collision is millions of times weaker and is completely drowned out by its own transmitter. A wireless sender literally cannot hear a collision while sending.
So collision detection is off the table, and wireless must instead try to AVOID collisions before they happen — a scheme called collision avoidance that the next guide unpacks in full. But avoidance has its own deep failure, and it is worth seeing the shape of it now. Picture three nodes in a line: laptop A on the left, the access point in the middle, laptop C on the right. A and C are both in range of the middle, but they are too far apart to hear each other. A listens, hears nothing from C (it cannot — C is out of range), and starts sending. C does the same. At the access point in the middle, their two signals collide and both are wrecked — yet neither A nor C ever heard the other to know it should have waited.
That is the hidden-terminal problem, and it is famous because it shows that "listen before you talk" is not enough when nodes cannot all hear one another. The fix is to ask the receiver, not the channel: a sender quietly requests permission to send, and the access point — which can hear everyone — grants it and tells the rest to hold off. That request-to-send / clear-to-send handshake, RTS/CTS style, is how Wi-Fi works around hidden terminals, and it is the heart of the next guide. For now, just hold the chain of reasoning: a radio cannot hear its own collisions, so detection is impossible, so wireless must avoid collisions, and even avoidance is hard because hidden senders cannot sense each other.
Why all this confuses the upper layers
Step back and collect the four hard truths of the air, because every later guide leans on them. Signals fade with distance and obstacles (attenuation), so range is short and unstable. Noise and interference never stop, so the usable rate rises and falls with the signal-to-noise ratio. Echoes pile up (multipath), so raw bits blur unless clever coding rescues them. And losses are common and bursty rather than rare, giving wireless a high, variable bit error rate. None of these is a defect in your equipment — they are the physics of sharing the open air.
Here is the sting in the tail, and a genuine misconception worth uprooting now. Classic TCP, which you met in the transport rung, reads packet loss as a signal of congestion — its congestion control assumes a lost packet means a router queue overflowed, so it brakes hard and slows down, like a careful driver lifting off the gas. On a wire that assumption is almost always right. On wireless it often misfires: a packet was lost not because the network is congested, but because a bit flipped in a fading, noisy, echo-filled channel. TCP brakes anyway, throttling a connection that had plenty of room. That mismatch — TCP misreading wireless loss as congestion — is the reason the last guide of this rung exists, and now you can see exactly why it happens.