Topic: Gold collapsing. Bitcoin UP. - page 201. (Read 2032248 times)

While I'm moderately against just increasing the block size. (I think that we should sort out the rest of the hard engineering infrastructure so we can actually support larger blocks first).

The main point is even on this forum we have nowhere near the 95%+ consensus that we need to do a hard-fork.

Gavin's latest proposal is just dangerous with only needing 75% of the miners to agree.

I would much prefer Bitcoin to continue to work with 1MB blocks, than be destroyed on the reckless quest for larger blocks.

So you've abandoned 20 MB? And Bitcoin XT? What made you change your mind?

I like Satoshi's quote on this:

http://www.mail-archive.com/cryptography%40metzdowd.com/msg09979.html


The nonchalance, to me, indicates a nice trust in markets to do their job of optimizing allocation over time. But - if Bitcoin ever starts to become a serious global economic force, mainstream economists are going to flip out over the above quote, given the initial-distribution algo didn't go through some deep analysis, etc... 

I think it's interesting how the engineering decision of making a simple/transparent (easy to implement, thus more secure) distribution algorithm trumped any potential detailed economic complexity, presumably due to Satoshi's understanding that the market would eventually allocate the capital optimally anyways, given the transparency of the current and future supply.

Those are tx merely being broadcast. "being capable of" means they can get into blocks at a sustained rate.

It would be cool to see that graph for transactions having 1 confirmation.

all over.  and it would be the same at 8MB:

hey, i thought the network was only capable of 2-3 tx/s?  look at our peaks today.  not bad:

i don't think any spam attacks can be sustained for a prolonged period of time.  it's too expensive as they pay normal tx fees.  while yes, at 1MB, they crowd out normal users, at 8MB or more, the spammers have to pay 8x more in fees.  while the anti Gavin crowd will say it's a drop in the bucket to perform such an attack even at 8x, what they i think they miss is that those tx fees will be tremendously beneficial for the miners raking up all those free fees and actually is expensive.  that will strengthen mining which is always a good thing for an industry that has been under stress for over the last year.  that would be the last thing they want to do.  can you also imagine a sustained 8x increase in the tx graph below?:



and what that would do to the price?  an 8x increase?  pt being, a spam attack is and will get more and more expensive and uneconomical and even strengthen mining.  it's be like your local candy store; owners ideally would like to sell to as many kids on the block as possible to keep them all happy but if one big kid keeps coming in week after week and buying out all your inventory, you aren't going to complain.  you might even prefer it as your turnover increases and you make more money.

all in all, this is just another form of manipulation of an open source, transparent system that won't work in the end and shouldn't be used as a reason for us not to adopt Gavin's plan.

edit:  yes, full nodes will be more expensive. but we already know they will become specialized.  and once pruning comes in, ordinary users can still run those.

There is no physical limit defined by nature, other than whatever the highest capacity node in the world can handle (which would be the only node able to keep up and therefore the network would consist of a single node).

In any other configuration the capacity is defined by the lowest capacity node on the network, which is not a limit defined by nature, but by intended or unintended consequences of the protocol design. If you want the lowest capacity node to be home computers, embedded devices, etc. then the protocol has to be designed differently than if you want that lowest capacity node to be a large or very large server in a data center. Nature allows either.

dying off again:

My guess is when we approach the limit, 1) people will learn what is the necessary fee to get their transactions into the first block, 2) the blocks will normally not be full for those transactions, 3) some people will pay lower fees and they are happy to risk waiting a few blocks sometimes. 4) The fee will be just high enough to discourage the least necessary transactions.

Said otherwise, since people's preferences are diversified, most people will be happy with the situation. This does not mean I want an artificial cap; the lower prices and the larger number of happy customers, the better chance bitcoin has to be the global money system everybody prefer. Only the physical limits of the system, as defined by nature, should exist.

what's interesting is that this attack is not slowing down block formation at_all.  we're just seeing continual full blocks.  based on Tradeblock's previous analysis, real customer tx's can probably get through if they pay a high enough fee.  unfortunately, the avg consumer won't know what that fee is given current wallet software capability and will get frustrated from delays.

edit:  altho unconf tx's aren't that bad at 4500

So you think his hardware is suffering from muscular atrophy?

no, my higher memory usage is coming from higher than normal unconf tx's.  the UTXO set is relatively flat at around 7MiB.  with bigger blocks, the unconf tx's could be cleared out much faster in bigger chunks.

fee rate dropping again:

unconf tx's are still only at 6400

nobody is expecting to not have to increase their full node capacity upon 8MB block sizes.  that's the price to pay to bring in new users.  as new users come in, more merchants will be born and host their own full nodes.   that's what we have been talking about as the network moves to more specialized full nodes.

The vested interest.

This is why I hate PoW mining as an economically significant paradigm.

ASIC hardware can't be repurposed.

I designed a far superior ASIC-resistant hash (meaning leverage the ASIC inside the CPU) than Monero's but I was still reticent because it is not sufficient to deal with that paradigmatic flaw, but then suddenly ugghh the epiphany and the rest is history waiting to be made.

This test demonstrates that cypherdoc's resources are at 25-45% of its capacity. And if we increase block size 8x then he can process only 2-4 MB blocks and his nodes will collapse.

Yes, the test demonstrates, if anything, that hell does not break lose when transaction pressure exceeds the limit. So it is not critical that the increase happens at any specific point. What we still don't know, is what happens when real people really want transactions to go through, and when the pressure keeps up over weeks. I think there is time to do it calmly.

increase load 20x and you will see what is chaos.

edit:
it is 1 MB block size limit what keeps your nodes happy. :-)