- From: Melvin Carvalho <melvincarvalho@gmail.com>
- Date: Tue, 5 May 2020 09:32:58 +0200
- To: Aaron Coburn <acoburn@apache.org>
- Cc: Martynas Jusevičius <martynas@atomgraph.com>, public-webid <public-webid@w3.org>
- Message-ID: <CAKaEYhKVNjC9FDTk-ZHF49tqqJ32m8RWEn5zAWumXFbyhSoMQw@mail.gmail.com>
On Tue, 5 May 2020 at 01:20, Aaron Coburn <acoburn@apache.org> wrote: > > I think all Solid servers use cookies > > No, not all Solid server use cookies for authentication. Some don't use > cookies at all, even when they use OIDC-based authentication. > To clarify, I was quoting from the chat with Dmitri (who did the auth for node-solid-server and made webid-oidc) Do you know which servers may use a cookie, and which not? That aside, do you think it's a good idea to document? > > > On Mon, 4 May 2020 at 18:49, Martynas Jusevičius <martynas@atomgraph.com> > wrote: > >> WebID-TLS does not require a cookie. >> >> On Tue, 5 May 2020 at 00.14, Melvin Carvalho <melvincarvalho@gmail.com> >> wrote: >> >>> Was chatting to Kingsley and Dmitri lately and I realized that quite >>> often WebID over HTTP is authenticated via a shared secret. Namely a >>> cookie. >>> >>> We realized that this is not documented anywhere. And I asked dmitri >>> about it usage. Im just capturing the responses here. >>> >>> "I think all Solid servers use cookies (and store the WebID in the >>> cookie session) as a local authentication method, in addition to TLS and >>> OIDC. It's very convenient; the only reason it's not the main mechanism is >>> of course, it's domain-specific." >>> >>> and >>> >>> "I don't think it's written up anywhere. I think it's because each >>> implementation's cookie session mechanism is slightly different. >>> But usually, it goes like: >>> >>> 1. When a user logs in (via username and password, or TLS, or OIDC >>> token), you put their WebID in your session. (request.session.webId = >>> <user's authenticated webId>) >>> 2. In all the other requests, you just use request.session.webId >>> directly. >>> >>> And the server's (Express.js, or whatever the other ones are using) >>> session cookie store takes care of it." >>> >>> This actually might be one of the more common types of WebID auth. >>> >>> Would it be worth writing up, and should it be called: >>> >>> WebID + Cookie or >>> WebID + Shared Secret? >>> >>
Received on Tuesday, 5 May 2020 07:33:24 UTC