> Hmm yes, this seems redundant in a way.
> However, I am strongly convinced that the reason to introduce the  
> HTML attribute was to give authors a "syntactically cleaner" way to  
> specify their character encoding.

I would say there were a couple of reasons:

1) documents may be served by protocols, e.g. direct file access, which
do not carry character set metadata.

2) most people learn HTML on servers where the ability to configure the
server is blocked for commercial, or, maybe, security reasons.  They 
continue to avoid congiguring servers, even when they do have full
access (although some people argue that, even then, departmental
politics mean that the server configuration and content are completely
divorced).