Open Id XRDS Discovery

Question

I am working with Open Id, just playing around making a class to interact / auth Open Id's on my site (in PHP). I know there are a few other Libraries (like RPX), but I want to use my own (its good to keep help better understand the protocol and whether its right for me).

The question I have relates to the Open Id discovery sequence. Basically I have reached the point where I am looking at using the XRDS doc to get the local identity (openid.identity) from the claimed identity (openid.claimed_id).

My question is, do I have to make a cURL request to get the XRDS Location (X-XRDS-location) and then make another cURL request to get the actual XRDS doc??

It seems like with a DUMB request I only make one cURL request and get the Open Id Server, but have to make two to use the XRDS Smart method. Just doesn't seem right, can anyone else give me some info.

Answer 1

To be complete, yes, your RP must HTTP GET on the URL the user gives you, and then search for an XRDS document reference and if found do another HTTP GET from there. Keep in mind that the XRDS may be hosted on a different server, so don't code up anything that would require the connection to be the same between the two requests since it might not be the same connection.

If in your initial HTTP GET request you include the HTTP header:

Accept: application/xrds+xml

Then the page MAY respond immediately with the XRDS document rather than an HTML document that you have to parse for an XRDS link. You'll be able to detect that this has occurred by checking the HTTP response header for application/xrds+xml in its Content-Type header. This is an optimization so that RPs don't typically have to make that second HTTP GET call -- but you can't rely on it happening.

Answer 2

The best advice I can give you, is to try to abstract your HTTP requesting a little bit, and then just go through the entire process of doing an HTTP request twice.

You can keep your curl instances around if you want to speed things up using persistent connections, but that may or may not be want you want.

I hope this helps, and good luck.. OpenID is one of the most bulky and convoluted web standards I've come across since WebDAV =)

Evert

Answer 3

I know I'm late to the game here, but I think you should also check out the webfinger protocol. It takes the standard "email as userid" pattern and lets you do a lookup from there to discover openid etc.