Gregory's ServerThis is a rather interesting read: https://bengo.is/blogging/2024-10-03-the-challenge-of-activitypub-data-portability/
 9 comments
> I think a lot of people want 'Account Portability' because what they really want is Single Sign On. This may be true, and hopefully all the work I've been doing around pushing Mastodon towards a more standardised OAuth system that borrows elements from OIDC helps here. For instance, I've just done the pull request for a userinfo endpoint, which would more easily enable this "single sign onβ approach. I think something @bengo does kind of miss in this article is how these decisions would impact end users, would it simplify things or add more complexity? βWait, now you mean I need to choose an authentication server AND a mastodon server???β For non-technical people, a lot of these proposals would absolutely loose them. Only technical minded folks even know that public/private keys are used in Mastodon.  @thisismissem hey I didn't miss that. FWIW I actually don't think new users should have an 'identity server' at all (until they want one). but we do agree new users shouldn't have to make the choice as you present it. longer discussion I'll write up later. @thisismissem this is what I was getting at with: @bengo aah, okay. The article was pretty technical, and for a lot of users they just want to get setup and start posting.. and then stuff happens later and their like "ughβ @thisismissem my goal is to get it so users can get setup and start creating posts without talking to a server (identity or social) *at all* and only replicate to one or more servers when they are ready/able/connected to share those posts with others. i.e. https://www.inkandswitch.com/local-first/  Hi @bengo @thisismissem, And each delivery must be mutually visible from the internet for the http signature callback. |
> There's no reason an ActivityPub server should demand to control the end-user's private keys.
Whilst I agree in principle, in practice, management of security keys is a right pain in the ass for end-users. Sure, you could do authentication via a PAKE (OPAQUE / SRP6a), and then derive a key-encryption-key from the users' password, but that introduces a lot of complexity.
If a user looses their security keys, then they can never continue, there is no password reset option there.