Just seven months after crying wolf that the current 1MB blocks would be insufficient for true Bitcoin adoption, true blockchain teckmologee success, and global community use of the currency for daily purchases, USGavin is backpeddaling at full speed and with the vigour of a man half his age. Where once he proposed that 20MB blocks would be oh-so superior to the current blocksize, y'know, except for the detrimental security consequences and the assuredness of economic ruin, he is now accepting the error of his ways, at least in this one regard.
Undeterred as this entirely ineffectual asset of the existing fiat regime is, USGavin has now set his sights on the Unspent Transaction Output (UTXO) Database as an avenue from which to "fix" Bitcoin. His most recent article, not-so-inexplicably written on the "ninja" platform, presents his concerns thusly :
This is the technical objection that I’m most worried about:
More transactions means more memory for the UTXO database
I wasn’t worried about it yesterday, because I hadn’t looked at the trends. I am worried about it today.
Those with longer memories will be understandably skeptical of anything that USGavin is presently worried about. If the blocksize didn't end up needing to be larger (au contraire !), why should we pay his ramblings any mind now, much less those he meditated on for all of a single sunrise and sunset ?
Looking more closely into USGavin's back-of-napkin calculations, we see that his prices are quoted in fiat, which is essentially irrelevant to the discussion of node costs. If we look at the following numbers and re-price them in BTC, there's nothing to say that the price of memory needed to run a node will increase. As we may well find out, the price of memory could very well decrease as the infinitely elastic ball X hits the infinitely heavy wall.
Today, the UTXO database is about 650MiB. DRAM costs about $10 per GB, so today if you have to spend about $7 on DRAM for your full node worst case, if you want absolute fastest access to the UTXO. No big deal. Assuming the UTXO set continues to double and RAM prices continue to drop 20% per year, next year you’ll have to spend about $11. Ten years from now, almost $800
This is not to say that bandwidth and hard drive costs aren't worth considering, just that USGavin's new focus on memory shows how badly he's been trumped in the field of battle of matters of consequence. No word yet on whether he plans to follow-up this particular charade with a focus on making Bitcoin nodes with more attractive cases. Y'know, for the ladies.