Future site of a pending but imminent storm over the definition of "relational". Get your ThreadMess goggles ready...

''No need for a storm, or even a breeze.  Wikipedia has a fine definition.  See http://en.wikipedia.org/wiki/Relational_model''

''All the ThreadMess this topic needs has already been, uh, messed up at http://en.wikipedia.org/wiki/Talk:Relational_model''

Based on that, it does not appear that only highly normalized tables qualify. Normalization techniques were an add-on to Codd's original ideas anyhow.

''It is not a matter of "qualification", and the higher levels of normalisation were a result of further research.  However, with each level of normalisation it is proven as to what anomalies are eliminated.  The higher the level, the fewer the anomalies.  Only failure to achieve 1NF would be considered in violation of the RelationalModel per se, and even that is subject to some controversy.''

* There are sometimes tradeoffs for ridding such "anomalies". Anomaly is in the eye of the beholder. It can also be called "flexibility" in some cases. Anyhow, we already have topics that debate the practical merits of high normalization. What I am concerned about here are the minimum rules to qualify as "relational". This is '''not a merit topic'''.

* ''I certainly agree that there are sometimes tradeoffs.  No question.  However, "anomaly" is definitely '''not''' in the eye of the beholder.  The anomalies caused by failing to achieve a given normal form are well-specified and unambiguous.  There is nothing "flexible" about an anomaly.  There *might*, however, be some questionable "flexibility" (I use the term very loosely here) to be achieved by, say, sticking with BCNF instead of proceeding to 5NF when 5NF is warranted, but at the expense of implementing additional application functionality to avoid the anomaly that 5NF prevents.  I can't think of any theoretical circumstance where that would be considered acceptable, but I suppose it could happen under some pathological condition -- such as being unable to change a 3NF or BCNF schema (it's used by a closed-source off-the-shelf tool, for example), or a client-side application that is profoundly limited in some respect -- or if the involved tables can be guaranteed never to be updated.'' -- DaveVoorhis

Wikipedia is not a good reference - it contains some valid info, but some incorrect (bunch of monkeys on typewriters).

Some interesting info about the relational model:
* http://ycmi.med.yale.edu/nadkarni/db_course/normalization_frame.htm
* RelationalModel

''Indeed, Wikipedia is not always an accurate source, but the summary of the RelationalModel is sound.  More comprehensive sources are, of course, AnIntroductionToDatabaseSystems and TheThirdManifesto.''

And if you come back in a month, it may be all different.

''Possible, but unlikely.  Technical subjects tend to be fairly well monitored.  It's biographical information that tends toward bogosity, as it's considerably more difficult to verify than established technical matters.  However, if you're concerned, and are looking for the original -- and arguably definitive and fairly minimal (fewer relational algebra operators are possible) -- definition of the RelationalModel, I refer you to http://portal.acm.org/citation.cfm?id=362685''
-------
See also: DoesRelationalRequireTypes