The genesis of RDF and RDFS is somewhat blurry because the discussions that led to the design decisions behind the 2 specifications are mostly hidden to the public. The first RDF working group's mailing list is a member-only list, which means that it is only visible to someone who has a member account at the W3C.
I have access to this part of W3 history, and I can say that Jeen's account of why there are two namespaces is rather inaccurate (update: he deleted his answer, see comments). The idea of having a schema language, or some other ways of specifying RDF schemas or vocabularies, existed from the very beginning of the RDF working group. In fact, a working group for the RDF data model and syntax started with the idea that a separate working group will deal with the schema-related spec. An RDF Schema working group started just 2 months after the RDF Model & Syntax one.
Apparently, the decision to split the RDF/RDFS namespaces in 2 was deliberate, with some things being considered integral part of the data model (e.g., properties, containers, reification) while others were specific to schema descriptions. In an email from the 24th February 1998 (not public), a certain Ron Daniel Jr. argues in favour of keeping separate namespaces for RDF model & syntax and for RDF Schema, but the email also says that it would be possible to use the same namespace with a different set of terms for the core model and for RDFS.
In August 1998, a public comment on the early drafts of RDF Model & Syntax and RDF Schema said "The set of terms that define RDF should not be split into the RDF and RDFS schemas" but it seems that this comment was never addressed.
Later in the early 2000s, when the second RDF working group laboured on a cleaner model, the distinction between terms that are core to the data model, and terms that are integral part of RDF schema became mostly irrelevant. Now we keep the two namespaces for backward compatibility and we have to explain everyone that they exist for historical reasons.
Thanks for the info! I wasn't aware of the whole debate behind these namespaces! I wanted to know why the subclasses rdf:Bag, rdf:Alt, rdf:Seq are in rdf namespace, while the superclass rdf: Container is rdfs namespace. I thought it'd be more reasonable to have them defined in the same namespaces. But if we think of the hierarchy (superclass, subclass) as the schema, it could make sense.
Good digging, I stand corrected (just goes to show how unreliable personal memory is - at least in my case :)). I've removed my answer.
Note that there was a discussion on the W3C Semantic Web mailing list where more history about the two namespaces was given by Dan Brickley, who was part of the first RDF working group.