This is how it should happen in any version of SQL server or any SQL database system. This is because SQL is declarative, not procedural. The difference is in the declarative, you're saying "put what's in column 1 into column 2 and what's in column 2 into column 1", whereas in the procedural you'd say "put what's in column 1 into column 2 and then what's in column 2 into column 1". A very tiny difference, but what a difference!

As for Carlos' example with a variable, that's an example of a quirky update, which I first learned of from here: http://www.sqlservercentral.com/articles/T-SQL/68467/ (which references http://www.simple-talk.com/sql/learn-sql-server/robyn-pages-sql-server-cursor-workbench/, which appears to be the original source of the term.)

Suffice to say it's a side-effect of the way Microsoft implemented a declarative language (SQL) within a procedural system (x86 software) -- one which works (when programmed meticulously under very specific circumstances), but isn't supported or guaranteed to continue to work in future iterations.