FxA Sync iframe flow #2101
FxA Sync iframe flow #2101
Comments
|
Can we get more info about the use case? Sync is pretty specific to the browser, supporting signin to Sync from arbitrary sites seems like it could be a big ball of yuck. |
|
Only from the stub installer, the onboarding tour, or authorized sites. I had a discussion with Michael Verdi this morning and realized how offering account registration during UI tours is not terribly helpful. |
|
|
|
This could be implemented by the FxAccounts.jsm component in the browser opening up a generic WebChannel to listen to messages from Open questions:
|
|
This comment has been moved to the checklist in the description. |
We may have bikeshedded this already, so appologies if I just don't remember it. But does it make sense to include some sort of verb in the URL for all oauth redirects and reserve I get a vague something-not-quite-right feeling about having the oauth redirect dance land at the root of the site, although I don't feel strongly about it and don't want to derail what you've got going on here. (Also we'll need an issue in fxa-oauth-server to adjust the meaning of the |
I had the same thought last night. The firstrun/Sync flow is not going to be an OAuth relier, so sending the user to |
|
There is only one item in this list that does not have a PR, #2791. I'm going to close this out as "this is ready enough to ship." |
|
\o/ |
So that I can connect my Firefox with Sync from more places, as a Firefox user, I want to be able to authenticate (register or login) from an iframe.
stomlinson's list of sub work:
A bi-directional WebChannel is needed to communicate with the browser. (#2231, PR #2358)An AuthenticationBroker that can communicate with Sync using the bi-directional WebChannel (#2359, PR #2237)A "Chromeless" UI is needed - one without the X and Firefox logo (#2198, PR #2236)The content server must allow/to be iframed. The iframe's parent origin needs to be checked to ensure we trust the framer. The OAuth flow uses theclient_id, then checks the parent origin is the same as is registered for theclient_id. This is not an OAuth flow. (#2242, PR #2364)The content server should decide whether to show/signin,/signup, or/settings- this can be done by visiting/. Expected behavior needs clarification. (#2197)Determine the URL API - how does the relier indicate:Chromeless UIAre in an iframe (OAuth reliers use context=iframe)Should use the WebChannel to communicate (webChannelId=sync possibly?)Is using Sync (currently we use service=sync&context=fx_desktop_v1) and no "signout" link should be displayed on settings.A possibility: https://accounts.firefox.com/?service=sync&context=iframeUnlike the normal Sync flow, the "settings" page should be displayed after sign in, and the "account verified" page should be displayed in the original tab after account verification. (#2674, PR #2687)TOS/PP rendering results in a scroll bar in the center of the page - should TOS/PP agreements be loaded in a new tab? YES! (#2351, PR #2643)Links in iframed TOS/PP replace content - (#1968, PR #2337)Channel timeout needs to go - (#2146, PR #2358)Disable the browser's "remember password" after signin/signup. (#2373, PR #2374, https://bugzilla.mozilla.org/show_bug.cgi?id=1173688)Remove the title for sync/iframe (#2565, PR #2593)Passcampaignandentrypointto the verification page (#2491, PR #2688)PassuniqueUserIdto the verification page to tie together signup/verification metrics (#2491, PR #2756)Determine whether the user is part of metrics in a more deterministic fashion (#2491, #2780)Notify the parent when the content height changes (#2566, PR #2594)Possibly consume GA query parameters and report them to metrics (#2579, PR #2689, PR #2779)Send consumed GA stats to Google (#2782, PR #2787)Prevent sign out link from showing on the settings page for accounts authenticated from firstrun (#2658, PR #2853)Fix partially hidden body (#2669, PR #2675)Notify iframe parent on initial screen load, login, and verification (#2704, PR #2705)Fix padding around marketing snippet on verification page (#2709, PR #2798)send a signal to the parent on successful signup (#2790, PR #2793)Bugzilla tracking bug: https://bugzilla.mozilla.org/show_bug.cgi?id=1150231
The text was updated successfully, but these errors were encountered: