I didnt understand the answer of why a 5 or 6 minute block would be bad, other than a very small increase in orphan rate and delaying any txfee auction market for a while
The answer was that the asymmetry is such that making the time too short (given network parameters, which are not really measurable and not stable) is much, much worse than making it too long.
I continue to assert that reducing the time should not be confused with increasing capacity. A block time reduction by a factor of two,
all else being equal, would also be accompanied by a reduction in the block size by a factor of two.
Looking at the data, it seems the negatives of 5 minute block are pretty small and that would give 2MB of blockspace per 10 minutes. So if we assume the error rate difference between 5 and 10 minutes is small enough to being considered equal enough, then having 2 1MB blocks in a 10 minute period sure seems like double the tx capacity per 10 minutes.
Not sure why the blocksize is reduced at half the time, that would defeat the purpose.
given the assumption that 5 and 10 minutes are equal enough, then why would the blocksize need to be cut in half? We can also assume that 1MB per 5 minutes is not any problem for the nodes' bandwidth and wouldnt increase the error rates appreciably.
What error rate at 5 minutes requires cutting the size in half. If there is no need to cut the size in half, then how is it not a doubling in capacity?
Confused...
James
To answer your last question, if you increase the capacity then you are doubling the bandwidth and storage requirements. If you want to do that, and you have buy-in to do that from the relevant stakeholders, then you might as well just increase the block size.
What you are are ignoring in "the data" is this:
The scaling of this depends on factors like network hashpower distribution and latencies which are hard to measure and which change.
Since 5 minutes is "reasonably close" to the inflection point on the graph where being too low becomes
much worse and being too high becomes
only slightly worse, it seems quite reasonable in light of the above quote to err on the side of (possibly) too high and just implement a 2x capacity increase with an increase in block size rather than a reduction in block time.
In all honesty, and lacking in hard data, I do
believe that 5 minutes would be fine, but what's the point? Going well below 5 minutes would most likely not be fine, so that limits any available capacity increase via that method to roughly 2x. Increasing the block size has no such limit.
thanks. mostly makes sense, but the last part sounds like doubling via 5 minutes is a onetime gain and we will have to increase blocksize at some point anyway, so no sense in doubling via 5 minutes.
using that logic, it leaves a "free" doubling unused and to me that seems a waste. Also 5 minutes would offer the advantage of half the latency, which many would view as a positive. so it is not just a lack of negative, but a presence of positive
James
P.S. There are ways to increase capacity 10x or more without increasing blocksize or even reducing the blocktime, but would require more significant changes in the core, but doing both would boost the overall capacity and I thought more capacity was better than less