But the spirit is right. Just do it on a real computer at home on a landline. A static website is easy, incredibly safe (compared to, say, browsing with JS turned on), and useful.
Moreover, even if everybody is a scrooge who can't just eat a negligible annual cost without worrying about it, you don't need everybody. You just need a few extra nodes to provide redundancy for you. Which could be your sister and your best friend and your father in law rather than some strangers you have to pay in a way that requires you to file weird tax paperwork.
You have a cafe/small business owner who already owns a smartphone: that device has a decent chunk of data plan allowance, CPU and memory.
They'd like a small site to promote their business online; so they install the web server app and it allows them to add, edit and publish basic content - a few images, some blog posts, etc.
Self-hosting like this means they don't have to sign up for a monthly subscription, their users aren't subject to any tracking (assuming there's none embedded in the content generated by the app), and service provision scales until they need to upgrade to a dedicated host.
The questions about radio and battery usage are totally valid, so perhaps those need more consideration (user superkuh has some suggestions there).
In short: the resources and capacity already exist and we've already paid for them. The question is whether we can re-use those rather than being sold more of the same.
Imagine a collection of independent storage pools managed by various entities that control access, e.g. the IT guy of your family, your employer, some free software collective, some small business association. You join one (or more than one) based on a referral or personal relationship or whatever criteria they want to use. Then you have an account with a quota in excess of what any ordinary person is going to require but which doesn't allow Bob to dump 500TB of garbage into the pool without first negotiating with the administrators.
At which point the altruistic donations, i.e. the idle capacity of the devices of everyone in the pool, are enough to handle everybody's non-abusive workloads.
Discoverability and routing is certainly a question too. Something like ngrok could help there, with good network sandboxing on the user's local device.
There'd no doubt be times when the site would become popular enough that a single smartphone wouldn't be suitable and load-balancing the requests could be required.
ngrok only works because there's a company willing to proxy someone's traffic for free. In that sense "using ngrok" doesn't solve the issue, it only pushes the problem to someone else.
Sometimes you have to make do with what's available currently while preparing the future possibilities.
Phones are not servers.
Even ignoring this issue, it would open your phone up to getting DDoSed by strangers on the internet.
The only real solution is selling little boxes for less than 50 dollars that you can plug into your router and make it dead simple to host a website without any technical knowledge.
I've run some internal-facing websites on a Raspberry Pi, also a low-cost system, but have never tried publishing them on the wider web.
Yet HN never misses the opportunity to dogpile on anyone who points this out: data trasfer costs, battery usage, jokes about the Silicon Valley sitcom.
As if most phones weren't connected to wifi and power for 10+ hours a day.
I don't know of a single US mobile network that provides publicly addressable IP space, v4 or v6, it's all behind more than one layer of NAT.
AT&T doesn't even allow direct internet access, every single connection goes through their horrendously bad transparent proxy that breaks TLS 1.3, breaks ESNI, causes a SSL_RX_RECORD_TOO_LONG connection reset if you negotiate only modern ciphers, and sits in the middle of every connection to prevent things like obfsproxy.
[0] = https://ipfs.io/ [1] = https://scuttlebutt.nz/
An interesting notion, but would only work until someone's site became very popular. Then it all goes pear-shaped. And I expect that most people would want their sites to be very popular, or it wouldn't be worth their time.
There's a lot of unused computing power on phones that are charging and doing nothing overnight. If only there was a way to harness that computing power and forward the results once a day or something.
The business models of website/page-builder platforms are based on gathering a long tail of small businesses, and then selling the audience attention via ad networks and data mining.
Looking at the ecosystem from a distance, it's creating financial value, but there are technical means to provide the same actual business value in cheaper and more privacy-preserving ways.
(By "actual business value" I mean that a person can view the menu at a local cafe in their area -- a valuable interaction for both parties -- without that lookup being intermediated by a platform and N different third parties where a kind of "shadow market value" is extracted)