I think the point that the previous posters were making is that pure REST should not rely on out-of-band information. The client should not have to guess or create URLs from knowledge they may have to look up in some other place. A pure REST API should be discoverable and subsequently navigable without having to rely on any other information not contained inside the request.

In practice this can increase the amount of requests required to get at the information you want, and also the size of each request, as it must then include information about the actions that can be taken on the resource. It could be argued that this constraint is more for the benefit of systems or programs that need to interact with the API without knowing anything but the initial URL, and not necessarily a programmer who can work these things out. owever, if a REST API uses this constraint, and application developers follow it, deep changes can be made to the URL structure of the API without breaking the applications that use it.

It's not so much that it was inaccurate (although it was) as that it was missing the point.