What are the downsides to 8MB blocks?

[Dire]

This is a good question, I've also wanted to clearly know the answer to this.

What about companies that have API's or have built something connected to the blockchain? Will they have to do something at the fork? All of them? Because I imagine that would be a nightmare to do every two years.

Or is it not like that?

An increase in size of the blocks changes nothing to the nodes and daemons, only, of course, in the main mining code.

Ah, thanks for the answer.

After reading quite a lot of stuff about the block size, and in particular a post by Litecoin guy, I have to say, I can't see what the exact point is to keep the blocks at 1mb. I mean, in the long run Bitcoin will stand to lose more than it will gain by not adapting.

I'm no miner so I know I basically don't know, but as far as I can understand it, Bitcoin needs bigger blocks, and even if it needs bigger blocks in the future and another fork, then so what?

PC hardware was never designed to be upgraded in the first place, but it was built on... and on... and on. It wasn't the end of the world, quite the opposite in fact.

I remember when I was using a 1k 'computer'. You couldn't even fart with 1k nowadays.

[RoadTrain]

I have to say, I can't see what the exact point is to keep the blocks at 1mb.

Neither can I. I believe we need a gradual rise which is compatible with technology improvement. 15-20% a year looks reasonable to me. XT's proposal doesn't.

[iCEBREAKER]

My admittedly very limited understanding: Currently because of the limited block sizes the bitcoin network could never handle the traffic if a large amount of new bitcoin users came online and started using bitcoin. The network would be so slow that people would not have a very good experience.

The solution to this is to increase the block size to 8 MB per block allowing for many more transactions and a much faster network capable of handling an event in which a large amount of new users start using bitcoin.

I don't know of the downside though. Why are so many people against this? I don't understand.

8MB doesn't allow for "for many more transactions" only ~8 times the current (extremely low) number.

You can't scale by adding more of the same.  More of the same risks negative marginal returns, which is why that approach is called "bloating."

I'm not sure if you are aware of the centralization pressure that occurs when miners are allowed to pick their own block sizes. Miners have incentives to encourage a very high orphan rate, if the orphan rate is properly targeted. Mining is a 0-sum game, which means that miners have incentive to hurt eachother. (less competition means lower difficulty and more blocks for you).

The ideal scenario for a miner is that the block they produce immediately hits a large portion of the hash power (at least 51%, but I'm not sure what the optimal percentage is), and then never hits the remaining hashpower at all. That remaining hashpower will have to find 2 blocks before the rest of the hashpower finds a single block - this will be detrimental to the profits of the minority hashpower, significantly so.

Large blocks provide a way to achieve this to some effect for bigger miners....

...This attack can be levied by a single large miner with no collusion. Simply by prioritizing the other large miners, and refusing to send the block to the low-bandwidth minority, a big miner can create blocks large enough to heavily impact the orphan rate of smaller miners without significantly affecting their own orphan rate.

This is very bad. The block size limit is important to stop this class of attack.

The vast majority of research demonstrates that blocksize does matter, blocksize caps are required to secure the network, and large blocks are a centralizing pressure.

Here’s a short list of what has been published so far:

1) No blocksize cap and no minimum fee leads to catastrophic breakage as miners chase marginal 0 fees:

    http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2400519

It’s important to note that mandatory minimum fees could simply be rebated out-of-band, which would lead to the same problems.

2) a) Large mining pools make strategies other than honest mining more profitable:

    http://www.cs.cornell.edu/~ie53/publications/btcProcArXiv.pdf

2) b) In the presence of latency, some alternative selfish strategy exists that is more profitable at any mining pool size. The larger the latency, the greater the selfish mining benefit:

    http://arxiv.org/pdf/1507.06183v1.pdf

3) Mining simulations run by Pieter Wuille shows that well-connected peers making a majority of the hashing power have an advantage over less-connected ones, earning more profits per hash. Larger blocks even further favor these well-connected peers. This gets even worse as we shift from block subsidy to fee based reward :

    http://www.mail-archive.com/bitcoin-development@lists.sourceforge.net/msg08161.html

4) Other point(s):

If there is no blocksize cap, a miner should simply snipe the fees from the latest block and try to stale that block by mining their own replacement. You get all the fees plus any more from new transactions. Full blocks gives less reward for doing so, since you have to choose which transactions to include. https://www.reddit.com/r/Bitcoin/comments/3fpuld/a_transaction_fee_market_exists_without_a_block/ctqxkq6

[Andy4.4]

Downsides......  8mb block take much time to spread throughout the bitcoin network or block chain.

There are many places in world where Internet is costly and speed is also not good.
So according to me, this is a downside.

[LiteCoinGuy]

no bandwidth problem. no storage problem.

i would like to see mass adoption. i would like to follow satoshis view of big blocks. i would like to have a permission less and cheap blockchain (cheap transactions) for everybody in the world.


look at Nielsen's Law:

http://www.nngroup.com/articles/law-of-bandwidth/

we dont talk about 100 MB blocks tomorrow - we talk about 6, 8 or 16 MB blocks in a few years (and it takes time to even fill them).
hundreds of millions of people can support such nodes today. no problem.

[iCEBREAKER]

no bandwidth problem. no storage problem.

i would like to see mass adoption. i would like to follow satoshis view of big blocks. i would like to have a permission less and cheap blockchain (cheap transactions) for everybody in the world.


look at Nielsen's Law:

http://www.nngroup.com/articles/law-of-bandwidth/

we dont talk about 100 MB blocks tomorrow - we talk about 6, 8 or 16 MB blocks in a few years (and it takes time to even fill them).
hundreds of millions of people can support such nodes today. no problem.

Nielsen's (Folk) Law is bounded by Gibson's Razor:

"The future is already here - it's just not evenly distributed."

See my lengthy, well-researched response to you here for Important Details.   

[Soros Shorts]

Storage and internet bandwidth are not really an issue these days. The real issue is the CPU resources used during block validation. Every new block that a node receives has to be validated against the known blockchain. This includes blocks that a new node downloads for the very first time. Some idiots are still running nodes on Raspberry Pis or Beagle Bones. When/if we move to 8MB, these nodes will not be able to cope with the larger blocks in a reasonable amount of time.

If you run a p2pool node and bitcoin node on the same machine (like I did), you already see a problem today when 1MB blocks come in and you are running anything less than a 4-core current generation CPU. My solution was to run p2pool and bitcoind on separate i7 machines, which was a pain in the ass but my setup can now handle 1MB blocks without too many orphans of p2pool shares.

[LiteCoinGuy]

no bandwidth problem. no storage problem.

i would like to see mass adoption. i would like to follow satoshis view of big blocks. i would like to have a permission less and cheap blockchain (cheap transactions) for everybody in the world.


look at Nielsen's Law:

http://www.nngroup.com/articles/law-of-bandwidth/

we dont talk about 100 MB blocks tomorrow - we talk about 6, 8 or 16 MB blocks in a few years (and it takes time to even fill them).
hundreds of millions of people can support such nodes today. no problem.

Nielsen's (Folk) Law is bounded by Gibson's Razor:

"The future is already here - it's just not evenly distributed."

See my lengthy, well-researched response to you here for Important Details.  

i read it. but even when you look Rustys estimation, bigger blocks are no problem:

Extracting the figures gives:

#Average download speed in November 2008 was 3.6Mbit
#Average download speed in November 2014 was 22.8Mbit

#Average upload speed in November 2008 to April 2009 was 0.43Mbit/s
#Average upload speed in November 2014 was 2.9Mbit


http://rusty.ozlabs.org/?p=551


@Soros Shorts

that could happen. but please note that blocks will probably not be full immediately.

[fonsie]

What's the point of a "speedy 1MB blocks flashing by over 10gbps uplinks" bitcoin network, if nobody can use it.

[andysbizz]

This will help to solve the spam attack, and send more transactions to the network

[Kazimir]

Some idiots are still running nodes on Raspberry Pis or Beagle Bones.

Exactly what is so idiotic about supporting the Bitcoin network by running dedicated devices as nodes (like Raspberry Pis) ?

[Brewins]

Some idiots are still running nodes on Raspberry Pis or Beagle Bones.

Exactly what is so idiotic about supporting the Bitcoin network by running dedicated devices as nodes (like Raspberry Pis) ?

they will soon run out of disk space and I don't think raspeberries are supposed to handle so much disk accesses, so it might reduce the life of the device

[cellard]

Sooner or later only a few people will be able to run nodes. And even if running nodes was easy, most people would be too lazy and resort to SPV or web based wallets. So the moral of the story is: No matter what they do, the % of people running nodes is always going to be a minority of enthusiasts.

[fonsie]

Some idiots are still running nodes on Raspberry Pis or Beagle Bones.

Exactly what is so idiotic about supporting the Bitcoin network by running dedicated devices as nodes (like Raspberry Pis) ?

they will soon run out of disk space and I don't think raspeberries are supposed to handle so much disk accesses, so it might reduce the life of the device

A 1TB drive attached to a Raspberry Pie doesn't run out of diskspace sooner then if it were attached to a Core i7.
I'm also not sure why it would die from "disk accesses".

[Linuld]

8MB per 10 minutes is about 13.6 KB/s. What was the problem again?

Not that I think that simply increasing the block size limit to 8MB is a good solution (I'd rather see something more dynamic) but if 13.6 KB/s is actually a problem, maybe you should fix your internet, rather than debate over Bitcoin.

Miners cant wait 10 minutes to have their newly mined block trasfered to the next node.

What do you mean with that? Miners can propagate really fast. An there is a project, i'm not sure anymore what it's name was, that was created to propagate blocks faster. And they get >50% in 2 seconds or so when using that network. My numbers might be wrong but it was so fast that creating 1 transaction blocks made no real sense.

[Linuld]

The biggest downside is basically the fact that the blocksize will be bigger, therefore limiting the amount of people that can access to it, or they could still access to it but it will be slower to download it.

Not really. Where should all these additional transactions come from suddenly to fill the 8MB? The blocks weren't full up to the spam attacks so why should suddenly there be so much more transactions?

If you think about spammers then no, that would simply be way too expensive.

[Linuld]

It will solve the spam attacks problem.

How is that? What prevents anyone from sending 8 times (or 100 times) as much spam as what we've seen recently?

What we need is a slight change in the default fee setting for Bitcoin nodes, that still allow very cheap payments (also micropayments) but helps against mass spam / dust transactions. Something similar to what Litecoin does, by requiring an additional small fee for each small output in a tx (and perhaps for each small input as well, while we're at it).

Will you do it? You would be poor pretty fast. Even the small KB that were needed to fell the 1MB were already expensive. Imagining that someone would spend the money to fill 8 MB blocks is hard.

It simply would not make any sense in any way to spam that way.

[uxgpf]

There are some problems with bigger blocks, as far as I can understand bigger blocks means:

1. You need to have more HDD place to store bigger blockchain.

Partly true. If you're running a node you can prune the blockchain. I run with a node with 10GB blockchain currently, but even pruning it to 512MB is possible.

2. Transactions fees will be actually higher, RIP bitcoin microtransactions.

False. Transaction fees will be lower, because there's more space in the block, thus less fee competition.

3. Higher blocks also requires higher internet bandwidth.

True. Now running a node requires some 5kB/s, so running a node won't be a problem for a long time. Propagation times for miners might be, though.

[uxgpf]

Some idiots are still running nodes on Raspberry Pis or Beagle Bones.

Exactly what is so idiotic about supporting the Bitcoin network by running dedicated devices as nodes (like Raspberry Pis) ?

they will soon run out of disk space and I don't think raspeberries are supposed to handle so much disk accesses, so it might reduce the life of the device

Just use some external drive and prune the blockchain to size of your liking. I haven't had any problems running node on raspberry pi and won't have until bandwidth of my ca. $25 connection is not enough (would need over 100X increase of blocksize from today). Currently upload bandwitdh used is under 5kB/s for most of time (of which some might be other traffic as I also run other services on it).

[Linuld]

There are some problems with bigger blocks, as far as I can understand bigger blocks means:

1. You need to have more HDD place to store bigger blockchain.
2. Transactions fees will be actually higher, RIP bitcoin microtransactions.
3. Higher blocks also requires higher internet bandwidth.

Point 1 is not really the case. I mean the 8MB won't be filled magically suddenly. Where should these transactions come from. And when you would stay at 1MB only because of that then bitcoin would be unuseable since ALWAYS transactions would not be included. What kind of transaction system would it be when you can't trust that it transfers?

Point 2 is wrong. The fee will grow without a bigger blocksize. With 1MB blocksize for example everyone would pay higher fees to get their transaction included before others.

Point 3 is the same like point 1.