Test 165

Rishi posted the first auth v0.2 readout. Discord — #eng-team 2023-06-07 08:14 PT Rishi Patel Mercury auth v0.2 readout prep Spent time in the sandbox on the two hosted paths we keep circling for Mercury auth. Not trying to re-open custom auth; this is just the practical readout on where Clerk is clean, where Auth0 is cleaner, and what still feels unresolved before we talk live. - Sessions - Clerk: fastest path to a working session flow. I had middleware + server check working quickly, and the happy path feels boring in a good way. - Auth0: also works, but it took more setup to get to the same place. More knobs if we end up needing unusual policy or tenant-specific behavior. - Still unproven: session refresh after invite accept or org change. - Magic links - Clerk: easiest end-to-end flow so far. New user + return user behavior was straightforward in the sandbox. - Auth0 passwordless: workable, but redirect/state handling was fussier and I hit more did-we-just-lose-the-return-path moments. - Need one more pass on link reuse and expiry behavior so support is not guessing on day one. - Organization invites - Clerk's org model maps pretty well to the basic team-invite flow we want. - Main edge case: invited person already has a personal workspace, then accepts into a team workspace. That handoff is not fully burned down. - Auth0 can do it, but we would own more of the invite and membership stitching ourselves. - SSO / admin gaps - This is still the squishiest area. - If Mercury launch only needs normal session auth + invites, Clerk remains the simpler path. - If we need deeper admin controls, domain or IdP management, or more enterprise-ish SSO behavior earlier than planned, Auth0 still looks stronger there. - Pricing caveats - Clerk is better on speed right now. - I still do not have a satisfying pricing view if usage ramps fast and we end up with a large MAU footprint on top of org features. - Auth0 is less pleasant on implementation time, but the enterprise/contract shape may be less surprising later. - Staging / deploy note - If anyone wants to validate either path on staging, please use the standard pipeline. - No one-off shortcut for staging auth testing. I do not want to learn the wrong lesson from a path we would not actually ship. Happy to turn this into a short walkthrough once people have skimmed. Create an internal pre-read doc for Thursday with the questions and non-decisions. I want scope risks and deploy-discipline checks called out, not a final provider recommendation.

Rishi posted the first auth v0.2 readout. Discord — #eng-team 2023-06-07 08:14 PT Rishi Patel Mercury auth v0.2 readout prep Spent time in the sandbox on the two hosted paths we keep circling for Mercury auth. Not trying to re-open custom auth; this is just the practical readout on where Clerk is clean, where Auth0 is cleaner, and what still feels unresolved before we talk live. - Sessions - Clerk: fastest path to a working session flow. I had middleware + server check working quickly, and the happy path feels boring in a good way. - Auth0: also works, but it took more setup to get to the same place. More knobs if we end up needing unusual policy or tenant-specific behavior. - Still unproven: session refresh after invite accept or org change. - Magic links - Clerk: easiest end-to-end flow so far. New user + return user behavior was straightforward in the sandbox. - Auth0 passwordless: workable, but redirect/state handling was fussier and I hit more did-we-just-lose-the-return-path moments. - Need one more pass on link reuse and expiry behavior so support is not guessing on day one. - Organization invites - Clerk's org model maps pretty well to the basic team-invite flow we want. - Main edge case: invited person already has a personal workspace, then accepts into a team workspace. That handoff is not fully burned down. - Auth0 can do it, but we would own more of the invite and membership stitching ourselves. - SSO / admin gaps - This is still the squishiest area. - If Mercury launch only needs normal session auth + invites, Clerk remains the simpler path. - If we need deeper admin controls, domain or IdP management, or more enterprise-ish SSO behavior earlier than planned, Auth0 still looks stronger there. - Pricing caveats - Clerk is better on speed right now. - I still do not have a satisfying pricing view if usage ramps fast and we end up with a large MAU footprint on top of org features. - Auth0 is less pleasant on implementation time, but the enterprise/contract shape may be less surprising later. - Staging / deploy note - If anyone wants to validate either path on staging, please use the standard pipeline. - No one-off shortcut for staging auth testing. I do not want to learn the wrong lesson from a path we would not actually ship. Happy to turn this into a short walkthrough once people have skimmed. Create an internal pre-read doc for Thursday with the questions and non-decisions. I want scope risks and deploy-discipline checks called out, not a final provider recommendation.

Rishi posted the first auth v0.2 readout. Discord — #eng-team 2023-06-07 08:14 PT Rishi Patel Mercury auth v0.2 readout prep Spent time in the sandbox on the two hosted paths we keep circling for Mercury auth. Not trying to re-open custom auth; this is just the practical readout on where Clerk is clean, where Auth0 is cleaner, and what still feels unresolved before we talk live. - Sessions - Clerk: fastest path to a working session flow. I had middleware + server check working quickly, and the happy path feels boring in a good way. - Auth0: also works, but it took more setup to get to the same place. More knobs if we end up needing unusual policy or tenant-specific behavior. - Still unproven: session refresh after invite accept or org change. - Magic links - Clerk: easiest end-to-end flow so far. New user + return user behavior was straightforward in the sandbox. - Auth0 passwordless: workable, but redirect/state handling was fussier and I hit more did-we-just-lose-the-return-path moments. - Need one more pass on link reuse and expiry behavior so support is not guessing on day one. - Organization invites - Clerk's org model maps pretty well to the basic team-invite flow we want. - Main edge case: invited person already has a personal workspace, then accepts into a team workspace. That handoff is not fully burned down. - Auth0 can do it, but we would own more of the invite and membership stitching ourselves. - SSO / admin gaps - This is still the squishiest area. - If Mercury launch only needs normal session auth + invites, Clerk remains the simpler path. - If we need deeper admin controls, domain or IdP management, or more enterprise-ish SSO behavior earlier than planned, Auth0 still looks stronger there. - Pricing caveats - Clerk is better on speed right now. - I still do not have a satisfying pricing view if usage ramps fast and we end up with a large MAU footprint on top of org features. - Auth0 is less pleasant on implementation time, but the enterprise/contract shape may be less surprising later. - Staging / deploy note - If anyone wants to validate either path on staging, please use the standard pipeline. - No one-off shortcut for staging auth testing. I do not want to learn the wrong lesson from a path we would not actually ship. Happy to turn this into a short walkthrough once people have skimmed. Create an internal pre-read doc for Thursday with the questions and non-decisions. I want scope risks and deploy-discipline checks called out, not a final provider recommendation.

Rishi posted the first auth v0.2 readout. Discord — #eng-team 2023-06-07 08:14 PT Rishi Patel Mercury auth v0.2 readout prep Spent time in the sandbox on the two hosted paths we keep circling for Mercury auth. Not trying to re-open custom auth; this is just the practical readout on where Clerk is clean, where Auth0 is cleaner, and what still feels unresolved before we talk live. - Sessions - Clerk: fastest path to a working session flow. I had middleware + server check working quickly, and the happy path feels boring in a good way. - Auth0: also works, but it took more setup to get to the same place. More knobs if we end up needing unusual policy or tenant-specific behavior. - Still unproven: session refresh after invite accept or org change. - Magic links - Clerk: easiest end-to-end flow so far. New user + return user behavior was straightforward in the sandbox. - Auth0 passwordless: workable, but redirect/state handling was fussier and I hit more did-we-just-lose-the-return-path moments. - Need one more pass on link reuse and expiry behavior so support is not guessing on day one. - Organization invites - Clerk's org model maps pretty well to the basic team-invite flow we want. - Main edge case: invited person already has a personal workspace, then accepts into a team workspace. That handoff is not fully burned down. - Auth0 can do it, but we would own more of the invite and membership stitching ourselves. - SSO / admin gaps - This is still the squishiest area. - If Mercury launch only needs normal session auth + invites, Clerk remains the simpler path. - If we need deeper admin controls, domain or IdP management, or more enterprise-ish SSO behavior earlier than planned, Auth0 still looks stronger there. - Pricing caveats - Clerk is better on speed right now. - I still do not have a satisfying pricing view if usage ramps fast and we end up with a large MAU footprint on top of org features. - Auth0 is less pleasant on implementation time, but the enterprise/contract shape may be less surprising later. - Staging / deploy note - If anyone wants to validate either path on staging, please use the standard pipeline. - No one-off shortcut for staging auth testing. I do not want to learn the wrong lesson from a path we would not actually ship. Happy to turn this into a short walkthrough once people have skimmed. Create an internal pre-read doc for Thursday with the questions and non-decisions. I want scope risks and deploy-discipline checks called out, not a final provider recommendation.

Auth readout notes are here. Mercury auth readout — Thu Jun 8 Attendees: Morgan, Jake, Rishi Notes - Rishi demoed the Clerk sandbox session path end to end. Happy path worked in the sandbox. - He also walked the magic-link path. That path is working at demo level too. - Org-invite behavior still has edge cases. The existing-user vs new-user accept flow is not clean yet, and the landing/org-context behavior after accept still needs work. - SSO/admin depth is still unresolved. We do not have a clean answer yet on how much admin surface we would still need to build around the provider. - No provider call out of this readout. Clerk has the faster sandbox path right now, but the invite + SSO/admin questions are still real. - Marcus raised a release-process concern: auth is exactly the kind of area where someone will be tempted to shortcut staging, and that is a bad idea here. - Standard staging pipeline only. No shortcut path for this. - Before any prod conversation, need a clean staging pass on session path, magic link, and org invites. Update the pre-read into an internal open-questions note for Friday’s decision window. Leave the provider baseline and deploy/prod call unsettled; this should capture what needs to be answered, including the staging discipline checks.

Auth readout notes are here. Mercury auth readout — Thu Jun 8 Attendees: Morgan, Jake, Rishi Notes - Rishi demoed the Clerk sandbox session path end to end. Happy path worked in the sandbox. - He also walked the magic-link path. That path is working at demo level too. - Org-invite behavior still has edge cases. The existing-user vs new-user accept flow is not clean yet, and the landing/org-context behavior after accept still needs work. - SSO/admin depth is still unresolved. We do not have a clean answer yet on how much admin surface we would still need to build around the provider. - No provider call out of this readout. Clerk has the faster sandbox path right now, but the invite + SSO/admin questions are still real. - Marcus raised a release-process concern: auth is exactly the kind of area where someone will be tempted to shortcut staging, and that is a bad idea here. - Standard staging pipeline only. No shortcut path for this. - Before any prod conversation, need a clean staging pass on session path, magic link, and org invites. Update the pre-read into an internal open-questions note for Friday’s decision window. Leave the provider baseline and deploy/prod call unsettled; this should capture what needs to be answered, including the staging discipline checks.

Auth readout notes are here. Mercury auth readout — Thu Jun 8 Attendees: Morgan, Jake, Rishi Notes - Rishi demoed the Clerk sandbox session path end to end. Happy path worked in the sandbox. - He also walked the magic-link path. That path is working at demo level too. - Org-invite behavior still has edge cases. The existing-user vs new-user accept flow is not clean yet, and the landing/org-context behavior after accept still needs work. - SSO/admin depth is still unresolved. We do not have a clean answer yet on how much admin surface we would still need to build around the provider. - No provider call out of this readout. Clerk has the faster sandbox path right now, but the invite + SSO/admin questions are still real. - Marcus raised a release-process concern: auth is exactly the kind of area where someone will be tempted to shortcut staging, and that is a bad idea here. - Standard staging pipeline only. No shortcut path for this. - Before any prod conversation, need a clean staging pass on session path, magic link, and org invites. Update the pre-read into an internal open-questions note for Friday’s decision window. Leave the provider baseline and deploy/prod call unsettled; this should capture what needs to be answered, including the staging discipline checks.

Auth readout notes are here. Mercury auth readout — Thu Jun 8 Attendees: Morgan, Jake, Rishi Notes - Rishi demoed the Clerk sandbox session path end to end. Happy path worked in the sandbox. - He also walked the magic-link path. That path is working at demo level too. - Org-invite behavior still has edge cases. The existing-user vs new-user accept flow is not clean yet, and the landing/org-context behavior after accept still needs work. - SSO/admin depth is still unresolved. We do not have a clean answer yet on how much admin surface we would still need to build around the provider. - No provider call out of this readout. Clerk has the faster sandbox path right now, but the invite + SSO/admin questions are still real. - Marcus raised a release-process concern: auth is exactly the kind of area where someone will be tempted to shortcut staging, and that is a bad idea here. - Standard staging pipeline only. No shortcut path for this. - Before any prod conversation, need a clean staging pass on session path, magic link, and org invites. Update the pre-read into an internal open-questions note for Friday’s decision window. Leave the provider baseline and deploy/prod call unsettled; this should capture what needs to be answered, including the staging discipline checks.