The Limitations of the Ethernet CRC and TCP/IP checksums for error detection (noahdavids.org)
kaif 積分 0 編輯於

So, if there is an undetectable corrupted segment on the network once every 34 hours or even once a month, why aren't the databases hopelessly corrupt? Why aren't applications randomly failing? Well, in many cases a corrupt segment is no big deal. For example, a bad ACK segment will be ignored; a corrupt byte or two in am image file will probably not be noticed, and no one would be surprised if a document I sent them had a spelling error.

kaif 積分 0

其實是在看QUIC怎麼處理error detection的... 看最新版本的RFC1 完全沒有提到, 連早期的Fowrard Error Correcting都沒有了