- From: Mark Nottingham <mnot@akamai.com>
- Date: Tue, 20 Mar 2001 07:43:22 -0800
- To: Yves Lafon <ylafon@w3.org>
- Cc: Henrik Frystyk Nielsen <frystyk@microsoft.com>, xml-dist-app@w3.org
A 'reverse proxy' is a gateway which uses the same protocol on both sides, more or less. A gateway is an intermediary which looks like a server, but satisfies requests through the use of another service on the other side. The difference between this kind of intermediary and a server is somewhat slippery; both look like an origin server to the client, while the surrogate (usually) uses a small bit of information (the identity of the true origin) to forward requests, and perhaps apply some processing to it. The line does get blurry, especially when the device composes the service. I think that to us, the interesting question is whether the message or part of it persists in some fashion through the device, so that we can apply targetting, etc. I'd also caution against drawing too much from the HTTP for modeling these relationships; we're going much further. Cheers, On Tue, Mar 20, 2001 at 01:30:08PM +0100, Yves Lafon wrote: > On Mon, 19 Mar 2001, Henrik Frystyk Nielsen wrote: > > > > > Yves, > > > > I think what you are describing in this scenario is typically called a > > gateway or in some cases a reverse proxy (a term that I don't really > > like). You are right that it is an intermediary but it really is an > > Well, yes and no, in the first transaction, S did act as a reverse proxy > (Mark, is reverse proxy used even if the protocol used are different on > both sids?) But in the second case, it will actually generate the real > reply according to the pieces received, it is then the real receiving XMLP > application. > If you want to keep it as an intermediary, then (as mentionned in 3.3) we > should allow it to modify/delete the payload. > > > In a previous version of [1] which you can find at [2]. Node IV is an > > example of an application layer intermediary. > > If you want to keep S (Nove IV) as an intermediary, then (as mentionned > in 3.3) we should allow it to modify/delete/regenerate the payload. > > -- > Yves Lafon - W3C > "Baroula que barouleras, au tiéu toujou t'entourneras." > -- Mark Nottingham, Research Scientist Akamai Technologies (San Mateo, CA USA)
Received on Tuesday, 20 March 2001 10:43:30 UTC