This should probably go in, though I'm not sure if this alone is visible enough.

@gmaxwell I checked and bitcoinj and libbitcoin don't have specific code to take a redeemScript and product a P2SH address/scriptPubKey. bitcoin-ruby and bitcoin-js do, but do not warn about the pushdata limit. (bitcoin-js's implementation is particularly nuts, and treats any unknown script as a P2SH) My python-bitcoinlib does, and is the only one that checks for excess length.

So yeah, I wouldn't be at all surprised if someone gets caught up by this eventually and loses some funds due to poorly tested code. Might be worth it to re-write the BIP so that it first describes the point of P2SH, followed by the implementation details implementors need to know. Similarly we should also mention how the coinbase "voting" code is now obsolete, and that P2SH checks are simply applied after a specific time. (or better yet, just point implementors to the Satoshi codebase for details on this rather than giving the wrong impression that the BIP document is the standard)

ACK