[Bitcoin-development] SatoshiDice and Near-term scalability

[Stefan Thomas]

I do agree that changing/lifting the block size limit is a hard fork
measure, but Mike raised the point and I do believe that whatever we
decide to do now will be informed by our long term plan as well. So I
think it is relevant to the discussion.

> Can someone please help quantify the situation?  kthanks :)

According to BlockChain.info we seem to have lots of small blocks of
0-50KB and some larger 200-300 KB blocks. So in terms of near term
measure one thing I would like to know is why miners (i.e. no miners at
all) are fully exhausting the available block size despite thousands of
transactions in the memory pool. I'm not too familiar with the default
inclusion rules, so that would certainly be interesting to understand.
There are probably some low hanging fruit here.

The fact that SatoshiDice is able to afford to pay 0.0005 BTC fees and
fill up the memory pool means that either users who care about speedy
confirmation have to pay higher fees, the average actual block size has
to go up or prioritization has to get smarter. If load increases more
then we need more of any of these three tendencies as well. (Note that
the last one is only a very limited fix, because as the high priority
transactions get confirmed faster, the low priority ones take even longer.)


On 6/15/2012 7:17 PM, Jeff Garzik wrote:
> On Fri, Jun 15, 2012 at 12:56 PM, Stefan Thomas <moon at justmoon.de> wrote:
>> The artificial limits like the block size limit are essentially putting
> [...]
>
> Changing the block size is an item for the hard-fork list.  The chance
> of the block size limit changing in the short term seems rather low...
>  it is a "nuclear option."
>
> Hard-fork requires a very high level of community buy-in, because it
> shuts out older clients who will simply refuse to consider >1MB blocks
> valid.
>
> Anything approaching that level of change would need some good, hard
> data indicating that SatoshiDice was shutting out the majority of
> other traffic.  Does anyone measure mainnet "normal tx" confirmation
> times on a regular basis?  Any other hard data?
>
> Clearly SatoshiDice is a heavy user of the network, but there is a
> vast difference between a good stress test and a network flood that is
> shutting out non-SD users.
>
> Can someone please help quantify the situation?  kthanks :)
>