Looking for other third-party SPs that fail with stricter SameSite settings
Brent Putman
putmanb at georgetown.edu
Thu Feb 6 18:48:45 EST 2020
On 2/6/20 5:07 PM, Cantor, Scott wrote:
>> Currently this shows that ServiceNow and Instructure have apparently now
>> been fixed by the vendors;
> Allowing that it may be rolling out to tenants, I can say that OSUs Canvas instance is not fixed. Last I spoke to people here, they were denying responsibility, but that was Tier 1 support and I suspect their developers are working on it.
I just retested Georgetown's Canvas. It does now seem to be working
correctly with Chrome 80 on my MacOS 10.14.6 desktop + the 2 SameSite
flags enabled, even when launched with
--enable-features=SameSiteDefaultChecksMethodRigorously, to disable the
2 minute grace window for the "Lax+POST" case. A ps confirms it is
effectively running with all 3 flags set:
--enable-features=CookiesWithoutSameSiteMustBeSecure,SameSiteByDefaultCookies,SameSiteDefaultChecksMethodRigorously
Curiously though, it still seems broken under Firefox with the two
similar SameSite-related options set in about:config. The failure is
that you seem to get in, but the URL you wind up in Canvas is
"/?login_success=1"; that query param is not typical. Then when you
click on a course in the Dashboard (or attempt to directly go to a
course URL), it winds up redirecting you back to the IdP and you wind up
back in the Canvas Dashboard in the same state. So you can't actually
get into anything beyond the initial Dashboard page.
So not sure what to make of that. Seems like Chrome and FF's behavior
have diverged. And so not entirely clear to me whether Chrome is really
truly working here - and will continue to work when Chrome is settled
into its final default state - vs is there some issue with the testing
methodology that is off (e.g. SameSiteDefaultChecksMethodRigorously
isn't doing what it claims).
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://shibboleth.net/pipermail/users/attachments/20200206/248a2966/attachment.html>
More information about the users
mailing list