Skip to content

Federated sessions (§ 3.3) orthogonal to your stated goals? #254

@maceip

Description

@maceip

the core spec creates per-origin keys. amazon.com gets one key, and google.com gets a different one. and because the keys are different, they can't be used like a shared ID to track. checks out....

however, the Federated DBSC section explicitly breaks this rule to allow key sharing.

worse? In Federated DBSC, a Relying Party (RP) can only use an existing hardware key if the Session Provider (SP?!) "allows" it in their /.well-known/ file. Google can simply refuse to add my startup to their relying_origins list....

my startup is forced to make users do a manual, high-friction login, while Google's own services (YouTube, Meet) stay "seamlessly secure." Google (the SP) effectively decides which small businesses (RPs) get to have a "smooth" user experience. It turns a security protocol into a competitive permit system.

the death of the "Open" Web and the rise of "Identity Blocs => to make Federated DBSC work, the RP and the SP have to mutually "handshake" via .well-known files. this creates a web where security is no longer a universal standard you just "enable," but a partnership you must negotiate. so we end up with Identity Blocs. You’re either in the "Google/Okta/Microsoft" garden where everything is secure and fast, or you’re in the "Open Web" where every site feels clunky and "less secure" because they don't have the "blessing" of a major Session Provider. It makes it nearly impossible for a new, independent identity system to gain traction.
Oh shoot

how do we get back to "Marking HTTP As Non-Secure" and less dragging legacy idps

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions