This article introduces a novel approach (in my experience) to tackling an old problem, and I think the concept behind it is great. However, the article could have been much better written, with greatly improved examples.

First, I would have liked to have seen a table in a sample database used, rather than [INFORMATION_SCHEMA].columns. This confused me at first.

Second, in the "grail" query, the second "row_number" column is aliased as "totrows". This should have been named something like rseq, since it actually represented the reverse sequence number, not the total number of rows. This was confusing too.

Third, I could not find mention of the fact that the columns in the OVER() clause of the "grail" query MUST include the primary key, and the columns must be listed in the same order in both OVER() clauses or it throws the calculation off.

Fourth, a lot of confusion could have been eliminated, and it would have been a lot more useful to the reader, if the article had also included some query examples with filter criteria to demonstrate the effect on the row count (viz. that # of rows returned was not the number of rows in the entire table).

Lastly, and this is a minor point, I was not familiar with the WITH construct. Speaking only for myself, it would have been nice if this had been explained too.

I give 5 stars for the concept, and 2 stars for the writing.