This has nothing to do with block size, but is an inherent property of PoW and economies of scale.  It is true that bigger blocks would push a bit faster to optimise the network topology to the reality of the client server nature of bitcoin but this is inverting cause and effect.  The network topology is not the cause of centralization, but its effect.  The real centralisation is the fact that only a few entities produce the data, the block chain.  The network only adapts to that reality but is not its cause.  PoW with specialized hardware and the economies of scale automatically lead to a few central mining centres.   That is not the block size or the network that push to that, it is the economies of scale in PoW and the smoothing out of the "lottery" (instead of winning once a whole block per year (or not), you prefer to have regular income with a pool).

This whole debate is misguided.  Block size has not much to do with it.  The real centralization is the fact that only 5 entities have majority hash power (and hash power is voting power in a PoW system), and that 14 entities have almost all hash power.  This is already the case.  The ILLUSION that the nodes are manifold, while they are only CLIENTS DOWNLOADING DATA from these 14 data centres through their proxies in the P2P network gives you the illusion that bitcoin is not centralized.  So this whole debate is simply about keeping up an illusion of decentralization by avoiding the network to adapt to the reality of the data flow.

No, really, once you have a cartel of pools, which is unavoidable with a PoW type income lottery, the network topology doesn't really matter, in *reality* you have a server/client system, where the pools are the collective server of block chain because they are the principal source of the data.  The other nodes in the network are nothing else but proxy servers of this data and ultimately, the clients.

The real P2P nature of the network only made sense if a block could be mined *anywhere* on the network ; if essentially all nodes were also miners.   Then the whole network is at the same time client and server, and the P2P style network is justified.
But if you have a specialized set of nodes that *make* the data, and a *large quantity* of nodes that only *use* the data, then you get an automatic adaptation of the network topology to the reality of the data flow: server/client.

Note that *amongst* the mining pools, the P2P topology still makes sense ; this is why they most probably connect in an almost full mesh, because they don't trust one another (subcartels of selfish miners might form).  But FROM the mining pools to the non-mining nodes, the relationship is not P2P, but is client/server, and hence the optimal network topology adapts to that given.

No, this is unavoidable for ANY form of PoW system in the long run.  It IS already the case BTW: 14 miner pools have essentially all the hash rate.

The reason is not "block size" or whatever.  The reason is the lottery of PoW, and the economies of scale.

Solo mining is not done much any more, because with solo mining, you win ONE BLOCK every two years or so.  That's too much of a lottery.

If you don't want more than 10% income fluctuation (RMS value) in 1 week of your income, it means that you must be part of a team that "wins 100 times" during a week.  As there are 1000 blocks in 1 week, you must hence be part of a pool that has 10% of the total hash rate.  --> there can be only 10 such pools !

Even if you accept larger fluctuations of income, there will at most be a few tens of mining pools.

Now, these mining pools need good network connections, to their miners, and to other mining pools, because every second lost is a second of hash rate lost.  As mining pools don't trust one another, they want to get good links to SEVERAL of their competitors, to avoid the possibility of "selfish mining" which needs variable network delays to get your private block in front of the public block.  

So, AUTOMATICALLY, this ecosystem will evolve towards "a few tens of pools with very good data connections and big data centres".

As an owner of mining gear, you have all interest to be in a big pool ; but you don't want pools the become monopolies, because then they will start eating off your fees.  So as an owner of mining gear, you are going to be such that you want to have "a few big pools".  In order for your mining gear to be efficiently used, you want to be able to have a good data connection to your host pool --> they need good data centres with good connections to all of their miners.

Once that is the case, automatically the above topology follows.  Has not much to do with block size.  Is intrinsic to the PoW system with specialized hardware (ASICS).  It was built into bitcoin from the start.

You make this overly complicated.  Replay attacks are impossible from the moment you mix with newly mined coins on the split chain.  No need to make it so complicated.

Look, after the first "whales" block is mined, the coinbase of that block is split in dust (you'll have to wait for many blocks before you are able to spend the coinbase, but that's not a problem), and each of these dust UTXO can be used in mixers to split for good any coin, making a replay attack of that spending impossible on the other fork, because exactly that coinbase transaction simply didn't exist on the other fork, the "miner fork" (other miner, other address, other signature !).  The signatures of both transactions are hence totally different.  The transaction will be different and hence the signature will be different, and will not be valid on the other fork --> no replay possible.

No need to have to fully block the small blocks in order not to spend your coin in a replay attack.  Simply mix with dust from a new coinbase transaction.  Exchanges should quickly get themselves some of that dust ; big whales will find a way to get some of that dust too.  And eventually, mixers can be set up for the lesser life forms, mixing in some dust.

--> you need more sleep

Actually, it IS "downloading websites".  There are essentially 14 producers of block chain: the main pools.  With them, you have more than 80-90% of the hash rate.  If these pools are no idiots, they are linked amongst themselves by a high-speed back bone network in, preferably, if they don't trust one another, almost full mesh configuration.  These 14 pools' data centres are the principal servers of block chain.  They make the block chain and hence are the ones having it as primary source.  
 
Any direct connection of your full node to one of their data centres (their powerful "node" that can probably handle hundreds of thousands of connections) gets the block chain directly.  If you want to get the block chain through a P2P connection from another non-important node, you are just using that other node as a proxy server of the block chain, and not from the main servers (the 14 data centres).

I say 14, but in fact, 5 is enough: 5 miners have more than half of the hash rate, and they, for sure, are on a high speed backbone.  It is in their interest to set up very strong nodes, so that the rest of the full nodes can connect directly to them, and not through a clumsy and slow P2P network.

So essentially, the mining pool concentration makes automatically from the bitcoin network:

1) 1 central back bone on which the principal miner pools are connected amongst themselves (from 5 to 14 say)

2) mostly direct server-client links to these 5 - 14 data centres by all "seriouis" client nodes, in order to get the most reliable block chain directly from the source as quickly as possible

3) at the periphery, small amateur nodes connecting in P2P mode to the serious client nodes, if they don't manage to go to one of the main servers.

Bitcoin's mining architecture does not favour a P2P network, it favors a central backbone + spokes network.  Note that it is not a "single server" network: all important miner pools (5 - 14) are independent servers of the same block chain.  You can connect your 'serious node" in P2P mode to several miner data centres in order to be sure that none of them is cheating on you and giving you a "retarded" block chain.

As the block chain is sort of "cryptographically signed" with proof of work, the fact that the source is centralized doesn't matter too much in its quality ; nobody can fake a block chain without spending huge amounts of wasted hash rate.  The only thing that can happen is that you get a "late" block chain, but to avoid that, it is best to get it DIRECTLY from those that produce it: the main mining pool data centres, instead of letting it "drip through" the P2P network at snail pace.

If you consider that bitcoin is a "fiat-to-fiat" sending device, yes it is a double layer.   That is:

1) you have permissioned, banking-style access to bitcoin through exchange 1 (where you send fiat from your bank account to the exchange, to withdraw true bitcoins on chain).

2) you can send bitcoins permissionless to whomever you want

3) on the receiving side, the person needs permissionned, banking style conversion of his received bitcoin on an exchange, and back to his bank account

With LN, this becomes the following:

1) idem

2) this time, you have to ask the exchange to send bitcoins to you through your permissionned channel on your "LN bank" where you are connected to.

3) you have to ask your LN bank if you can send a permissionned payment to your destination (your LN bank can accept or refuse, ask fees, conditions, explanations, ....), that is to say, to your destination's permissioned LN bank

4) on the receiving side, the person will need to ask his permissioned LN bank to send the coins to his preferred permissioned exchange, convert to fiat, and get that fiat on his bank account.

So, what has disappeared in the LN story, is the permissionless sending of coins from sender to receiver.  You have to go through your fixed, permissioned LN bank, and so does your counter party (your "LN hub" to which you are attached until your LN link is sufficiently used that you can settle on chain).

Not really.  You would make a tiny litecoin, forked off bitcoin.  Why not use the bigger litecoin that is a copy of bitcoin, but with:
1) 4 times faster block times (so 4 times more transactions possible)
2) another PoW algorithm (Scrypt) ; for the rest, LTC is bitcoin's code
3) already segwit ready to activate ?

Why is Core imitating litecoin after all these years ?

This simple truth is hard to understand for some who thought that they had the power because they had downloaded a piece of free software from the internet and were running it on their old PC...

No, these are the greater fools that were needed to pay for the joke, but they have now, in the mean time.

Indeed.  You cannot enforce a soft fork with non-mining nodes.  You can just bring them to a standstill.

At that point, the only option for the user is to downgrade to an earlier version of core, or to download a bitcoin protocol compatible piece of software somewhere else.

However, it is true that if you need to reward people by organizing a competitive lottery for maintaining consensus, then tail emission is the only potential solution if you want a fluid system.   The reason is that spontaneous fees are never a good way, because OR you have to need scarcity of transactions (like bitcoin is having now), which leads to a "reserve currency" status of the system ; OR you will have a fluid system with so small fees that the system becomes unsafe or unstable and cannot reach consensus.

The BU proposal that network lag is going to lead to self-regulation is stupid, because the only network regulation you will obtain is a fully blocked network.  In fact, you need tail emission if you are to reward those securing the system.  Fees alone is not stable.

The cycle of money

Without the dream of being rich 10 years from now because you could buy 10 coins which kept this bubble going, it will deflate and fall back on real economic value, which is minuscule.

However, once the eternal king  (also called god) is dead, everybody knows that the next king has a finite life time too.

Well, from the paper that describes this way of working:

https://www.dash.org/instantx/


--> the part in bold is a clear impossibility in a P2P network and is essentially the fundamental problem of consensus making.

and


Essentially, this mechanism proposes that once consensus is reached, it can reach consensus
If it were possible to ensure that "all nodes received the message", there would never ever be any consensus problem.  In other words, instant pay works well, if there is no double spend attempt.

After that, the miners still include what they want, if I understand well.  However, most probably they are asked to include instant-pay confirmed transactions first.

It is important to realize the circular argument in this, because it can sound credible.   The whole, and the SINGLE difficulty in setting up a crypto coin, and the exact reason why Satoshi invented Proof of Work and the Block chain, is that it is essentially impossible to:

1) make sure that EVERYONE on the network got the transaction

2) that the order in which two contradictory transactions are received on the network is the same for everyone

If one could make this happen, crypto would be MUCH MUCH simpler.  There wouldn't be any mining, blocks or whatever.  Indeed, every node on the network would receive ALL transactions in order, and only accept the first one if there's a double spend attempt.  There wouldn't be any need to put these transactions in blocks, to mine them, to "secure" them with PoW or anything of the kind.  That whole circus only serves ONE SINGLE PURPOSE: deciding, once and for all, in what order the propagated transactions should be processed, and picking (at random) one of the conflicting ones.

If you can SOLVE this problem in one way or another (like DASH pretends, with its propagating "locks"), then there's no need for blockchains, mining or any other securing.  Propagating the lock, or propagating the transaction, is the same.  The first one should be locking out the second one, and it is exactly because this cannot be rigorously imposed, that all the hassle with block chains has been invented.  So if you can propagate the LOCKS correctly, you could have propagated the transactions correctly.  And no more mining, blocks and whatever.

So instant pay solves the problem, when the problem didn't need to be solved.

1) bitcoin mining IS ALREADY centralized.  5 pools have more than 51%.  10 pools have more than 75%.

2) LN centralizes even faster than PoW.  Only big bank hubs can run the LN profitably.

Yes, I wonder what will happen to the "most secure chain" propaganda when Joe Sixpack can make a block on his PC.

If that is what people want, why don't they switch to litecoin ?  It is "bitcoin with another PoW", and with 4 times more room on the chain.  Moreover, it has Segwit also ready.  So the thing you are hoping for, exists already several years.  Why don't people switch massively to Litecoin ?

I will tell you: because it is called Litecoin.

Litecoin is almost as "pristine" as bitcoin.  The simple fact that it is not called "bitcoin" is what holds people back.  

This illustrates the irrationality of this whole happening.  Bickering over would-be things that exist already for years next door, because it doesn't have the right brand name.

What core is trying to do, is already done since years.  Next door.

If you "find a way to stop this", you've just rendered bitcoin not permissionless any more. 

This.  Because bitcoin is just a brand name now.

Uh, no !  DASH only confirms that your transaction is in the mem pool.  It also "confirms" that the mem pool won't allow another transaction double spend ; but that is fundamentally flawed because it would violate the CAP theorem. In other words, some master nodes may honestly confirm your transaction, while some OTHER masternodes may confirm your OTHER transaction during network delays, leading to an inconsistency in confirmed transactions.  The only way to avoid that, would be to require 100% masternode confirmation, at which point, THEY are the solvers of the consensus, and you don't need a block chain any more !  But it would be sufficient to have one badly behaving masternode and you would bring all of DASH to a stand still.  So this doesn't work.  Instant Pay only works at very low traffic rates.  It is a snake oil scheme of confirmation.
It still has to go on the block chain before really being confirmed, that block could just as well be orphaned, and gone is your transaction !

And why would miners appreciate transaction fees falling back to a penny ?
What silly miner is going to truly allow increase in block size if he can keep his peers at 1 MB and hence keep the fee market under pressure ?

What idiot miner is REALLY going to fork to kill the goose with the golden eggs ?