In case you missed it: http://www.sammur.com/?p=36
My reply, currently awaiting moderation: """ Hey, I think you are using an old version of Avro; in addition to C, Java, and Python implementations, there are Ruby and C++ implementations, and a C# one on the way (see https://issues.apache.org/jira/browse/AVRO-533)<https://issues.apache.org/jira/browse/AVRO-533%29> . Also, for maintainability, you don’t have to write your protocols or schemas in JSON; see http://avro.apache.org/docs/current/genavro.html for details and http://github.com/hammer/hbase-trunk-with-avro/blob/trunk/src/main/java/org/apache/hadoop/hbase/avro/hbase.genavrofor an example. Lastly, your point about type safety makes me think you are using the generic (no code generation, only use built-in data types) rather than specific (generate code, including objects corresponding to complex Avro types) implementation. You should be able to get compile-time type checks from the latter. In any case, Avro is certainly a less mature project than protocol buffers, but it’s also actively developed. Please submit any suggestions for improvements to the mailing lists (http://avro.apache.org/mailing_lists.html ). Regards, Jeff """ Please hop in if you have more thoughts to add. In any case, it certainly looks like we could improve the documentation. Thanks, Jeff
