thanks for your feedback. I have fixed the post accordingly. It is very good news for me that I was wrong and supercolumn names are indexed! It is really useful for denormalization. And I have already covered this demormalization use-case in my next post http://maxgrinev.com/2010/07/12/do-you-really-need-sql-to-do-it-all-in-cassandra/

This is an excellent post!

A couple clarifications:

- if you don’t have anything relevant to store in the column value, leaving it an empty byte array is fine

- it is _subcolumn_ names that are not indexed inside supercolumns, meaning that you should only store data inside supercolumns that you plan to access together. In your example the supercolumns would be fine. Top level columns (the supercolumn’s name) are always indexed.

- The most common use case for supercolumns is for denormalizing data from another columnfamily, e.g. in user_timelines you could make the tweet id a supercolumn name, with subcolumns of the actual tweet field names + values. This makes reads still more efficient, since you don’t have to perform joins manually via multiget at read time.