This.

The LN protocol can only lead to centralization of the LN network.
If we find it too centralized, we can use another network based on another protocol, such as Tumblebit, which also provides payment hubs and payment channels.

In any protocol stack, there are layers that are more centralized than others. Decentralization is not a state, it's a process.

That's certainly the sales pitch, the theory and the hope, but in practice, it's honestly too early to tell for sure.  There are questions around things like the expiration of a timelock if the transaction takes too long to confirm when attempting settling, which opens up the potential to counterparty theft.  Also considerations of how an infinite number of transactions back and forth without a fee might have a bearing on miner revenues over time as the block reward diminishes.  We also want to be certain that "layer 0" is still economically viable for the average person and that they aren't forced to transact off-chain due to excessive costs.  In something as complex and uncertain as LN, there are bound to be wide ranging implications.  That's not to say we shouldn't explore such potential, of course.  The benefits are absolutely worth it.  But at the same time, we shouldn't be too dismissive when people raise issues like this.

If I understand this correctly, most of these transactions will not be public? The payment hubs will process the tx's and then show the bundled result of that on the Blockchain? Who do you contact if something went wrong with your tx's and are these entities not a single point of failure, when they come under attack?

Many of us still need these types of questions answered, before we will switch. We will rely on miners fees in future, and we are uncertain if this will diminished our income, if this is implemented. ^hmmmmmm^

Analogy fail - computer networks aren't like a horse

To wit: the multiple protocols operating on top of the TCP/IP network that you used to post your incorrect statement to this website

As far as i believe their are at least 3 different projects working on a form of lighting,  which means users will have a choice as to which one to use.  Even Kim dotcom is creating Bitcache which is along the lines of some sort of Lightning network.  If this is the case then it wont be centralised as you have stated.

from the article:

Right now, I believe that altcoins are a viable alternative to solutions like Lightning, which are several years away. Bitcoin devs even say, "if you don't like the bitcoin fees, use something else". And in fact, that is what many people are doing. Using altcoins while exchanging to bitcoin also creates some anonimity. Services like Shapeshift are taking off, exchanges are supporting hundreds of alt cryptocurrencies, and their market caps are skyrocketing. Money is pouring into the alts as bitcoin becomes less and less useful, and the core devs sit there muttering about quadratic hashing with their thumbs up their asses...

I believe that layers built on top of bitcoin won't be adopted on a large scale.

The whole idea of second layers and in this case lightning network on top of bitcoin, is to guarantee that if the second layer  crumbles, the sturdy base (core of the network protocol) will remain decentralized. This is achieved by not having big blocks.

None of that makes any sense, because:

a) Lightning transactions aggregate the net amount of multiple Bitcoin transactions, on-chain settlement will never be 1:1 offchain:onchain, your scenario is exagerrated to the point of being ridiculous

b) You provided zero reasoning, just an entire paragraph of false assertions

Can't you read? I've answered this already, you're a time-waster