On Tue, Aug 18, 2009 at 3:20 PM, David Axmark<[email protected]> wrote: > On Tue, 2009-08-18 at 07:33 -0700, Jay Pipes wrote: > >> The Google Protobuffer message format I will propose is as follows: >> >> message Checksum >> { >> required uint32 checksum; /* crc32 value */ >> repeated string table; /* contains table_list or table_identifier */ >> } > > BTW, most resonable numbers we choose 15 years ago from MySQL did not > look that good 15 years later. > > And I am sure a 32 bit value here will look silly in 15 years. A 64/128 > bit will be as fast to compute. And if you have enough data one in 4 > billion may not sound as the best odds. >
My 2 cents: Allow checksumming partial data sets -- rows *and* columns. (there are times we've found non-matching data in a column, but wanted to continue checksums of the rest of the columns to ensure they're all in sync while we worked on fixing the non-matching data). Allow a pluggable algorithm, so that the algorithm can change. With mk-table-checksum, there are 2 algorithms besides the CHECKSUM, and one of them may give false positives; the other is not order dependent, so the rows (1,2) and (2,1) checksum to the same value. -- - Sheeri K. Cabral http://tinyurl.com/mysqlbook will take you to the Amazon.com page for my book, "MySQL Administrator's Bible". _______________________________________________ Mailing list: https://launchpad.net/~drizzle-discuss Post to : [email protected] Unsubscribe : https://launchpad.net/~drizzle-discuss More help : https://help.launchpad.net/ListHelp
