Incorrect `iss` value on tokens, not the same as `issuer` value in OpenID configuration

[michielbdejong]

When trying to log in to CSS using the run-against-css.sh script of the WAC tests, you get:

2021-08-11T12:45:33.820Z [DPoPWebIdExtractor] warn: Error verifying WebID via DPoP-bound access token: The access token issuer doesn't match its associated WebID's trusted OIDC issuers.
Actual: https://solidcommunity.net/
Expected: https://solidcommunity.net

So maybe we can change NSS to remove that trailing slash from the issuer?

[RubenVerborgh]

https://openid.net/specs/openid-connect-core-1_0.html#IssuerIdentifier reads:

The issuer returned by discovery MUST exactly match the value of iss in the ID Token. 

However, https://solidcommunity.net/.well-known/openid-configuration returns

"issuer":"https://solidcommunity.net"

so it is an NSS bug that it mints tokens with a different issuer.

[bourgeoa]

@jaxoncreed are you ok with the fact that it is a NSS bug.

[jaxoncreed]

Yeah, if there is no trailing slash in the openid-configuration, there shouldn't be one in the token.

[michielbdejong]

~~The value on https://solidcommunity.net/.well-known/openid-provider is loaded from /mnt/volume_lon1_01/.db/oidc/op/provider.json, I edited it in that file, and that fixed the direct issue for now. But we should also fix the code that (re-)generates this file when it is not there.~~

EDIT: this was a bad idea because it caused #1613 so undid this now.

[bourgeoa]

@michielbdejong So in fact you did not remove trailing slash to issuer but added a trailing slash.

"issuer" : "https://solidcommunity.net/"

[michielbdejong]

~~@bourgeoa yeah, and it broke something else so i'm trying to put it back now - can you pop into https://gitter.im/solid/solidcommunity.net ?~~

EDIT: fixed now, the server is back to how it was before last night

[michielbdejong]

I'm now trying out a less invasive work-around, namely hard-coding the issuer in ./node_modules/@solid/oidc-op/src/handlers/OpenIDConfigurationRequest.js

[michielbdejong]

I tried out various things but it also ties in with the issuer linked to from the webid profile, and apparently CSS isn't successfully extracting the issuer from the DPop token that the WAC tests send, so in the end I gave up trying to fix this situation and just commented out https://github.com/solid/access-token-verifier/blob/3f99fb7/src/algorithm/verifySolidAccessTokenIssuer.ts#L17 in my system-under-test. This issue still needs fixing though.

[RubenVerborgh]

apparently CSS isn't successfully extracting the issuer from the DPop token that the WAC tests send

Well, the token is there and CSS spells it out, so that can't really be the issue.

The issue is a bug in the test suite where https://solidtestsuite.solidcommunity.net/profile/card#me uses https://solidcommunity.net/ as issuer rather than the advertised https://solidcommunity.net. Because as you can see in https://github.com/solid/access-token-verifier/blob/3f99fb7/src/algorithm/verifySolidAccessTokenIssuer.ts#L9, it compares the advertised issuer against those listed in the WebID document, which should be exact matches according to https://solid.github.io/solid-oidc/#resource-access-validation.

This means that he test suite is actually testing whether NSS behaves incorrectly, and confirms that it's not a good idea to base a test suite on it.

[RubenVerborgh]

In related news, why is https://github.com/solid/web-access-control-tests/blob/main/run-against-css.sh publicly listing actual username/password combinations? Separately from the dozens of bad reasons I can think of, it basically means that anyone can break the test suite at any time. As tempted as I felt to fix the slash in https://solidtestsuite.solidcommunity.net/profile/card#me, the fact that I can does not seem good.