You could think of information resource as Web document, and of non-information resource as thing. (Not a definition, just so you get a rough idea.)
I tried to explain the difference in this answer:
a HTTP URI could identify the page itself OR the thing the page is about. You can't tell if an URI identifies the page or the thing by simply looking at it.
Example (in Turtle syntax):
<http://en.wikipedia.org/wiki/The_Lord_of_the_Rings> ex:author "John Doe"
This could mean that the HTML page with the URI
http://en.wikipedia.org/wiki/The_Lord_of_the_Rings
is authored by "John Doe". Or it could mean that the thing described by that HTML page (→ the novel) is authored by "John Doe".
It’s a good practice to make it possible that others can differentiate this when re-using your URIs or your data. However, this is not required; some think that this is a non-issue or not worth the effort.
The discussion about this differentiation is known as the httpRange-14 issue ("What is the range of the HTTP dereference function?").
Some possible solutions are:
- Hash URIs (see example in this answer)
- 303 URIs (see DBpedia example in this answer)
- Property categories (rather new and probably not well-known; see my question with a brief explanation)
So when you use "slash URIs", and you care about the differentiation (and I think you should!), then yes, you would have to use a 303 redirect.
Otherwise (if you’d send 200), all user agents that follow TAG’s resolution would think that your URI identifies an information resource.