According to the protocol specification at the Bitcoin wiki, each message header includes the first 4 bytes of sha256(sha256(payload) as a packet checksum. I suppose this checksum is used for packet validation, but I do not see any benefit of this, as Bitcoin utilizes TCP which already has (as far as I know) full protection from lost and damaged packets. So... why this checksum? Is it to maintain compatibility for future UDP clients?

"Full protection" is a relative term. TCP packets include only a simple checksum, effectively just the sum of all the 16-bit words in the packet. See RFC 793 section 3.1. This is a fairly weak integrity protection mechanism. The Bitcoin designers may have felt the use of sha256 gave better protection. It may also, as you say, have been intended to provide forward compatibility for running Bitcoin over UDP or other protocols. It may even just have been paranoia.

  • But this weak protection is good enough for other sensitive protocols like FTP and HTTP? They don't include checksums, right? (HTTPS/SFTP does of course, but just plain FTP/HTTP errors could be disastrous too). Thanks for answering though :) – Jori Mar 2 '14 at 15:04
  • @jori It's not necessarily good enough, but it's what we've got. – Nate Eldredge Mar 2 '14 at 15:15
  • I'm going to research the TCP checksum capabilities a bit further and then post back as I cannot believe the designers deliberately choose a bad schema for such an important task. It would be disastrous if using some executable file send across the network using FTP, resulted in unexpected.behavior. – Jori Mar 2 '14 at 15:26
  • @jori: Bear in mind that the design is almost 40 years old, and it was important that the checksum be easy to compute on primitive hardware. CRCs were probably considered but rejected as too expensive, and our modern cryptographic hashes had not been invented. – Nate Eldredge Mar 2 '14 at 15:39
  • 2
    Bitcoin originally didn't have this checksum, but Satoshi was having a lot of trouble with various garbled packets on the network, so he added it. It's better than the TCP checksum, and it also covers cases where bad data is sent to the socket. See the post here starting with "I'll take a look a the logs." – theymos Jul 31 '14 at 16:23

Your Answer

 

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.