Radio is analog, and modem converts analog signals to digital signals and vice versa.
Everything is analog, yet you can do digital things with the physics that is given to us. As you've noticed, when transmitting something digital over an analog channel, you typically use a modulator, and to receive the same you'd need a demodulator, and modem is just a portmanteau of these two words.
So is P2P sharing of digital files (.exe, .7z, .pdf or any other file) possible using radio spectrum?
Yes, and with Wifi, Cellular networks, and many other techniques, I'm sure you're already aware of the feasibility.
If yes then which tools do I need? Does WiFi work in frequency ranges other than 2.4 GHz, 3.6 GHz, 4.9 GHz, 5 GHz and 5.9 GHz? All these frequency ranges have lower range of propagation, a lower frequency would be needed for greater propagation range. Does WiFi work in MHz or in Hz range?
Wifi, indeed, makes a few assumptions on the channel (ie. a microwave channel with 20 or 40 MHz width will behave differently than the much narrower, much lower, much noisier) channels you can get at shortwave, medium and longwave ranges – how could you have 20 MHz of bandwidth if your center frequency isn't even that large?
So, no, Wifi itself can't work at these frequencies. Also, Wifi has to assume a bounded flight-time of signals to work efficiently. To top that up, large-area covering WiFi networks are a bad ideaTM, mainly because of the hidden station problem.
There's other techniques, but as soon as you think about it: Low-frequency spectrum simply is sparse because of the low frequency. Thus, to get sufficient bandwidth, you move to higher frequencies, which is why Cellular networks operate at 800 MHz and up. There simply was no spectrum available below.
All these frequency ranges have lower range of propagation
That's an alternative fact. There's numerous long-range and space-ground links at frequencies much higher than that.
So, yes, while Free Space Path Loss might be higher for increasing frequency, antennas of a given size also get higher gain – and that makes all the microwave links that are the backhaul of a lot of the high-rate communication done today feasible.
P2P means Peer-to-Peer – I'm going to take that literally: It's not impossible to build a mesh network out of microwave links (in fact, that's what people do that operate cell phone towers over areas, and TV, towers on remote hilltops) and route stuff through that.
There's WiFi-based mesh networking – you might want to talk to your friendly Freifunk enthusiast about B.A.T.M.A.N.; but I still think that for long-distance mesh networks with links that are point-to-point, you'd probably want to do with a simple frequency division microwave link (ie. one center frequency for A->B and a different one for B->A) rather than a collision-avoidance based packet network (which has no advantages in a pure point-to-point, fixed terrestrial application).
You can rather easily build such a single link using (proper¹) SDRs that allow you to send arbitrary waveforms by calculating these in your PC, and do the inverse for reception. Using a network interface in your SDR application would allow you to plug together an IP network.
If you want to have omnidirectional rather than point-to-point links, you must share a single piece of spectrum across all users – which inherently divides the possible channel capacity by the number of users. You'll find, with a bit of consideration, that the easiest way to make the most of the spectrum that's available, is if there's a central node that everyone agrees on as controlling access timing – that way, for example, transmissions can be scheduled without the need for any "dead periods" in which stations have to listen to avoid colliding with others. If you do that, you'd be building some kind of TDMA base station, much akin to GSM / GPRS / EDGE (basically, in ascending order of flexibility in which time can be allocated to the stations) base stations. If you add the ability of base stations to communicate, and for user equipment to change between these, you've built a cellular network. Congrats!
If you don't need high rate (and, as Phil suggested) can live with half-a-century-old technology, the AX.25 packet radio technology might be an option – but certainly not for sharing any substantial files. AX.25 itself is horrible on many levels; it is especially bad-suited for networks, as there's no good network layer that is commonly used. APRS is often used atop of that to build what can best be described as a low-rate, sparse, very suboptimal, chaotic-good aligned, networking/relay system. It reduces the possible channel utilization by not having any awareness of routes, and thus would immediately break down if a couple of stations tried to continously use the channels at max rate for any substantial amount of time (or data).
So, that's definitely not a candidate for file sharing.
¹I say "proper" because SDR
is often (in my opinion, mis-) used for devices that have a fixed functionality, but run the signal processing in fixed software. To be honest, under that nomenclature, every cheap Baofeng handset is an SDR.