cassandra - Difference between Thrift and CQL 3 Columns/Rows

Question

At the Cassandra Summit, it was mentioned that Thrift and CQL 3 have subtle differences in their definitions of columns and rows. The Google hasn't helped me understand this difference. The only information I can find is that the metadata is different, and as such, I shouldn't mix thrift and CQL.

1. What is the subtle difference (I've read a bit about metadata representations...)?
2. in what way does it break the compatibility?
3. Why is the change better?

I'm happy to read any docs that will help me, I've just been unable to find anything specifically related to this topic.

Answer 1

Here's the reading path I was taught to follow to answer this same question:

1. Intro to Apache Cassandra 1.2
2. A Thrift to CQL3 upgrade guide
3. CQL3 for Cassandra experts : DataStax
4. Does CQL support dynamic columns / wide rows?