That math is too simple, sorry.
It's intentionally simple to illustrate the point.
Network 1 would be a pure microtransactions network. In this case 100tx/sec would be too few for a global currency (It could be a regional cryptocurrency, but then it would not be a competitor to Network 2).
100tx/sec (a block size of ~15 MB assuming a 250 byte average 1-address-to-1-address-transaction) would be more suitable for a mid-size-transactions-network, let's say with an average transaction amount of $20. That would be 172 mn usd per day in transaction volume - that would be a more realistic competitor to the 1000-per-transaction chain (let's call it Network 1b).
You have to remember that high capacity (and consequent low fees) bring with them uses that are not necessarily economic in nature. For example, using the bitcoin blockchain as a distributed storage file, for storing all kind of irrelevant things - in a sense we are talking about txs with 0$ transfer of value. If someone wants to use the blockchain to store gigabytes per day in bullshit, paying peanuts, they can. And you can't stop them.
The question is now: would the costs to maintain network 1b prohibitively higher than that of network 2? If yes, then Network 2 could have a competitive advantage. Network 1b would mean a ~788GB/year blockchain. Even today that can be handled by consumer-level storage hardware - and more so in the near future.
Yes it can, but with decentralization cost and users relying on third parties.
Plus real usage is not even 1/10th of that, which means 9+/10ths of the blockchain would be occupied with crap. Why would anyone impose 700gb of crap on everyone else? It's insane.
And how did the value from Network 2 come from? That's the question.
By its users and its properties to be able to move around non-trivial amounts of money with very small fees.
The benefit of bitcoin widens the larger the amount of money being transferred is. If you want to transfer 10cents and have to pay 20 cents fee, that's not effective. If you want to transfer 100$ or 10k$ and have to pay 20 cents fee (saving you 3$ or 300$ from, say, a paypal) that's far better.
This is not something theoretical btw. You go over at apmex, for example:
http://www.apmex.com/product/102251/2017-1-oz-gold-american-eagle-bu?toppicks1And see:
Quantity Check/Wire CC/PayPal
1 - 9 $1,310.69 $1,365.30
+55$ for doing the exact same purchase through CC/Paypal...
A small-block network will have few use cases (as outlined in the link I brought in above), much less users and therefore is more vulnerable to speculation attacks because its "intrinsic value" (=usability / network effect) is much more limited and its liquidity is much lower.
I think the use cases are determined based on a profitability equation. Our threshold might be around 5-10$ right now, meaning that if you want to transact more than that, it's probably more profitable going through BTC, than through other electronic payment means. If paypal requires users to pay, say, 3-4% in fees + 0.35$, then a 10$ transfer would cost around 0.65-0.75$.
Are there use cases for moving around more than 5-10$ and saving money in the process, while also enjoying the benefit of knowing that no one will freeze your account, demonetize your currency, reverse your txs, etc etc? The answer is a qualified yes.
PS: Despite of all this I'm a Segwit and LN supporter and I don't want 15 MB blocks now, 2 MB with a 10-20% yearly increase (like Satoshi's idea) would be fine. But I want LN as a microtransactions network and not as the standard way to transact with BTC.
Keep in mind that in the same discussion, the previous message of satoshi, says:
We can phase in a change later if we get closer to needing it.
...which indicates he wanted blocksize increases on a need-to basis. Which is what I'd expect from a rational man who has glimpsed the problem of giving away too much space for abuse and seeks a solution. The automated proposal of his next message cannot evaluate needs. Humans can (subjectively). And that's where friction comes in.
