Re: [core] Call for reviews of draft-castellani-core-http-mapping

[Carsten Bormann <[e-mail]>]

On Mar 26, 2013, at 15:52, Floris Van den Abeele <[e-mail]> wrote:

> My problem lies with a cross protocol proxy (XPP) with the following interface http://www.proxy.com/.wellknown/core-translate/1.2.3.4_4567/foo/bar?a=3 (which is translated to coap://1.2.3.4:4567/foo/bar?a=3). This is also said to be an example of a reverse proxy (right?). However, in this case the HTTP client is aware that the XPP isn't the origin server,

Actually, it may not be -- this URI may have been supplied by a server to the client, and clients aren't supposed to be snooping into the structure of URI path anyway.

> the actual origin server is mentioned in the request, so I would say it resembles a forward XPP. Reading section 10.2 of the core coap spec seems to confirm this (one difference is that the coap scheme is omitted in the 2nd example, but it is implied anyway).

Defining a convention for how a reverse cross-proxy builds the URIs under which it offers the resources of an origin server might look like allowing a client to use the reverse cross-proxy as if it were a forward-proxy.  And that may actually be a benefit of such a convention, because it is hard to carry forward-proxy configuration information through client APIs (and, even more, through sequences of prepended proxies).  But I would expect these reverse proxies to make their URIs available through a resource directory, so there may never be a need for a client to compute a URI.  The convention would then just help in understanding what is going on, and in sending diagnostic requests to a reverse cross-proxy.

Grüße, Carsten