I'm guessing the "foreseeable future" would be that bitcoin and maybe a few other crypto's are as big as something like VISA or Mastercard and is used a long side fiat. Or is the foreseeable future not as big?
The tests Gavin ran wrt block size were for 20MB and 200MB. His results are that the *current* code (but with increased max block size and tested for block at max capacity) can be run on *current* (consumer, non specialized) hardware even at those sizes.
There's this meme that the network currently can handle around 7 transactions per second (tps). In reality, because of higher average tx size, it's below that (somewhere between 3 and 6 tps, from what I know).
All other things equal, a block size of 20MB would means something around 20 tps. 200MB, around 200 tps. Not bad, but still an order of magnitude short
of Visa's (reported)
2000 tps on average.
However, keep in mind: this is for unoptimized code. Gavin mentioned what he considers a number of "low hanging fruit" optimizations to the MIT crowd, among them blockchain pruning and vastly smaller block announcements.
So, to summarize:
(1) I'd say a reasonable target for stating positively that "Bitcoin is scaling up well" would be 2000 transactions per second (tps), the VISA number
(2) Current code, current hardware, bigger blocks gets us to ~200 tps, so about 10% of the goal
(3) There's a number of optimizations that seem to be widely considered technically possible (eg pruning), and the main limitation in implementing them seems to be a) time constraints, b) high security/testing requirements for live changes to the code, c) consensus issues, especially if changes require a (controlled) hardfork
(4) All of the above doesn't even factor in the assumed increase in available bandwidth to the consumer, and cheaper storage capacity. Even if Moore's Law (assuming exponential growth) won't hold forever, hardware and bandwidth will certainly continue to be improved, and become cheaper to employ.
So, point (2) means, if there'd be consensus for a controlled fork to change max block size to Gavin's proposal (20MB max now, then +40% per year), we could already scale up to about 10% of average VISA. The remaining factor 10 seems well within reach when consider points (3) and (4).
Beyond that (i.e. "the network handling *all* of humankinds transactions."), I have no idea if that's feasible, and I don't know if anyone calculated that though. What I had in mind with my original response in here was "Bitcoin scales to the size of the world's largest CC company", not "Bitcoin becomes the only money". I'd say, the former scenario wouldn't be half bad already.
Disclaimer: I'm providing a digest of what I understood from Gavin (and other devs') comments on this topic. There are probably more qualified people to answer this, user
DeathAndTaxes comes time mind. If there's any glaring technical mistake in my above post, please correct me on it.
