Quote from: TierNolan on March 19, 2013, 11:49:04 AM

Quote from: retep on March 18, 2013, 05:40:33 AM

That's not quite true. Version 1.0 didn't have a 1MB limit, but instead used the same 0x2000000 byte, or 32MiB limit, used for any serialized data. Satoshi later added MAX_BLOCK_SIZE so that miners wouldn't create blocks bigger than 1MB, but larger was still accepted.

That is interesting. It sets a "harder" max block size, even if the 1MB limit is increased.

To increase the limit beyond 32MB, it would be necessary to change the message structure, or allow blocks to be split over multiple messages.

Not really. The 32MiB limit on serialized messages is just an anti-DoS attack measure. There isn't anything inherent in the serialized messages themselves that limits them to 32MiB; to make >32MiB blocks just requires both numbers to be changed at once. That said you are correct in that allowing blocks to be split over multiple messages will happen, if only because sending a block can be faster because you can skip sending the transactions if the receiver already knows about them.

Quote from: TierNolan on March 19, 2013, 11:49:04 AM

Quote

The commit comment is deliberately misleading: "only accept transactions sent by IP address if -allowreceivebyip is specified"

Any reason to make them misleading? It would seem like a recipe to making alt-clients incompatible.

Satoshi was solidly against the development of alt-clients, even just to spend coins let alone mining. That said those were very early days and he probably wanted the flexibility to change things as Bitcoin's core evolved, and it did. There is some really kooky code from those days in the repository though; did you know Satoshi wrote the beginnings of an EBay like auction market and put it in the client itself?