Hi Rob, Thanks for the contribution. Marc Portier has also suggested an alternative design path that I quite like. See my comment the updated RFE: http://restlet.tigris.org/issues/show_bug.cgi?id=157
Best regards, Jerome PS: You should add yourself in the cc: field of the RFE > -----Message d'origine----- > De : Rob Heittman [mailto:[EMAIL PROTECTED] > Envoyé : mardi 21 août 2007 18:26 > À : [email protected] > Objet : Re: Shortcut local traffic between Restlet HTTP > clients and servers > > > > If you have time to help with the RFE, maybe by providing a > patch, it could > > still go into the upcoming 1.1 release :) > > I posted a fairly non-invasive patch and additional class to > the RFE. I'm sure you could improve upon it. > > http://restlet.tigris.org/issues/show_bug.cgi?id=157 > > It adds a LoopbackDispatcher class that is aware of both the > client and server routers in the component and can loopback > certain request roots only if explicitly told to do so by an > attribute set in the ComponentContext: > > List<String> loopbacks = new ArrayList<String>(); > loopbacks.add("https://localhost:8443/usrv";); > > component.getContext().getAttributes().put("loopback.list",loopbacks); > > This seems to work as one would expect, with the expected > performance gain and negligible cost. > > I explored down the road of attempting to transparently poll > server routes before dispatching to any client routes. (If a > valid HostRoute attached to the server router could handle > the request, prefer it to using a Client.) This doesn't work > well, because the host routes are generally greedy, and there > is probably a default host that will claim it can handle > anything, even things it shouldn't. I suspect that > transparently magical loopback behavior is going to cause > massive confusion for somebody at some point, especially if > their Restlets are running inside a larger container. I even > confused myself :-) > > This explicitly coded approach seems to do what I need it to > do, without breaking anything that currently exists ... thoughts? > > - Rob > > >
