Intent to Experiment: Cookies Having Independent Partitioned State (CHIPS)

[Dylan Cutler]

Contact emails

dylan...@google.com, kaust...@google.com 

Spec

https://github.com/WICG/CHIPS

Summary

Given that Chrome plans on obsoleting unpartitioned third-party cookies, we want to give developers the ability to use cookies in cross-site contexts that are partitioned by top-level site (or First-Party Set, where the site uses that feature) to meet use cases that are not cross-site tracking related (e.g. SaaS embeds, headless CMS, sandbox domains, etc.). In order to do so, we introduce a mechanism to opt-in to having their third-party cookies partitioned by top-level site using a new cookie attribute, Partitioned.

Link to “Intent to Prototype” blink-dev discussion

https://groups.google.com/a/chromium.org/g/blink-dev/c/hvMJ33kqHRo

Goals for experimentation

CHIPS is a new, opt-in technology meant to preserve a set of use cases (e.g. third-party embeds) that may break once third-party cookies are phased out while preventing cross-site tracking. We need to validate whether the proposed syntax and semantics solve these use cases prior to third-party cookie obsoletion by giving developers a way to test it in a scaled manner and provide early feedback. We hope to validate ergonomics, deployability, and backward compatibility. 

Experimental timeline

The experiment will start in M100 and run from March 31st, 2022 until June 30, 2022.

Any risks when the experiment finishes?

Since Chrome will not send and may delete partitioned cookies when it is started with the feature disabled, sites that set cookies with the Partitioned attribute during the experiment will no longer have those cookies available on clients' machines.

Reason this experiment is being extended

N/A

Ongoing technical constraints

None.

Debuggability

We have coordinated with the DevTools team to surface cookie partition keys to developers in DevTools. We have added a new cookie inclusion reason with a debug string when sites set Partitioned cookies incorrectly. We may also support surfacing partitioned cookies that are not included in requests because their partition key did not match the top-level site in DevTools.

Will this feature be supported on all five Blink platforms supported by Origin Trials (Windows, Mac, Linux, Chrome OS, and Android)?

Yes.

Link to entry on the feature dashboard

https://www.chromestatus.com/feature/5179189105786880

[Chris Harrelson]

LGTM

--
You received this message because you are subscribed to the Google Groups "blink-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAMCNMFRxpqxLUXAEfFqcACt-S7A8Y_6Q9is%3DCZJskiYuby0qJA%40mail.gmail.com.

[Yoav Weiss]

On Tue, Feb 1, 2022 at 8:14 PM 'Dylan Cutler' via blink-dev <blin...@chromium.org> wrote:

Contact emails

dylan...@google.com, kaust...@google.com 

Spec

https://github.com/WICG/CHIPS

Summary

Given that Chrome plans on obsoleting unpartitioned third-party cookies, we want to give developers the ability to use cookies in cross-site contexts that are partitioned by top-level site (or First-Party Set, where the site uses that feature) to meet use cases that are not cross-site tracking related (e.g. SaaS embeds, headless CMS, sandbox domains, etc.). In order to do so, we introduce a mechanism to opt-in to having their third-party cookies partitioned by top-level site using a new cookie attribute, Partitioned.

Will this be run as a third-party Origin Trial? As a Finch experiment? Other?

 

Link to “Intent to Prototype” blink-dev discussion

https://groups.google.com/a/chromium.org/g/blink-dev/c/hvMJ33kqHRo

Goals for experimentation

CHIPS is a new, opt-in technology meant to preserve a set of use cases (e.g. third-party embeds) that may break once third-party cookies are phased out while preventing cross-site tracking. We need to validate whether the proposed syntax and semantics solve these use cases prior to third-party cookie obsoletion by giving developers a way to test it in a scaled manner and provide early feedback. We hope to validate ergonomics, deployability, and backward compatibility. 

Experimental timeline

The experiment will start in M100 and run from March 31st, 2022 until June 30, 2022.

Any risks when the experiment finishes?

Since Chrome will not send and may delete partitioned cookies when it is started with the feature disabled, sites that set cookies with the Partitioned attribute during the experiment will no longer have those cookies available on clients' machines.

So, when the experiment finishes, sites that opted-in to that mode will lose their cookies and their users will e.g. be logged out, etc?

That seems like a deterrent. Is there a way around that? (e.g. migrate the cookies to the default 3P behavior when the experiment is done. Not sure how feasible that is..)

Reason this experiment is being extended

N/A

Ongoing technical constraints

None.

Debuggability

We have coordinated with the DevTools team to surface cookie partition keys to developers in DevTools. We have added a new cookie inclusion reason with a debug string when sites set Partitioned cookies incorrectly. We may also support surfacing partitioned cookies that are not included in requests because their partition key did not match the top-level site in DevTools.

Will this feature be supported on all five Blink platforms supported by Origin Trials (Windows, Mac, Linux, Chrome OS, and Android)?

Yes.

Link to entry on the feature dashboard

https://www.chromestatus.com/feature/5179189105786880

[Daniel Bratell]

Can you verify that I am getting this right.

1. The long term plan is to get rid of "tracking" cookies, or more specifically third party cookies shared between multiple first parties.

2. This will not change anything unless a site explicitly asks for their cookies to stop tracking people.

3. (outside this experiment) Eventually the default might change to "Partioned" and another flag will have to be used to keep tracking users cross sites.

4. (outside this experiment) Finally tracking cookies are disabled completely (similar to what Safari has done).

If that is right, should this prepare the syntax to allow for step 3, like having "Partitioned=Absolutely" and "Partitioned=Nope" instead of just partitioned?

In step 4 I assume "Partitioned" becomes a no-op since that is the only available stage?

Another question: Do you have partners ready to start testing this?

/Daniel

[Dylan Cutler]

Will this be run as a third-party Origin Trial? As a Finch experiment? Other?

This experiment will be run as a 3P Origin Trial, 

So, when the experiment finishes, sites that opted-in to that mode will lose their cookies and their users will e.g. be logged out, etc?
That seems like a deterrent. Is there a way around that? (e.g. migrate the cookies to the default 3P behavior when the experiment is done. Not sure how feasible that is..)

The reasoning behind why we didn't do that is that partitioned cookies allow the existence of multiple cookies with the same host key, name, and path to exist in separate partitions. Rather than coalescing these into one cookie (which one is the right one to keep, after all?), we decided to just remove partitioned cookies from clients' machines when the feature is disabled to provide deterministic behavior.

 The long term plan is to get rid of "tracking" cookies, or more specifically third party cookies shared between multiple first parties.

Correct.

 This will not change anything unless a site explicitly asks for their cookies to stop tracking people.

In the short term, yes, clients with partitioned cookies enabled will support both partitioned and unpartitioned cross-site cookies. Once 3PCs are removed (see link above) then only partitioned cookies will be allowed in cross-party contexts.

Eventually the default might change to "Partioned" and another flag will have to be used to keep tracking users cross sites... In step 4 I assume "Partitioned" becomes a no-op since that is the only available stage?

I imagine when we first turn off 3PC that third parties will still need to explicitly opt into using partitioned state using the Partitioned attribute. If third parties do not opt into this behavior then they will be unable to use cookies at all. But, in the long term, we may have the Partitioned behavior be the default for cross-site cookies. In that case, the Partitioned attribute could just be ignored and eventually deprecated.

 

If that is right, should this prepare the syntax to allow for step 3, like having "Partitioned=Absolutely" and "Partitioned=Nope" instead of just partitioned?

I don't think we need the Partitioned attribute to have any other semantic meaning other than a flag saying "I am opting into receiving partitioned 3P state", so we decided to design it like the Secure and HttpOnly attributes (i.e. its presence in the cookie line being "true", it's absence being "false").

Do you have partners ready to start testing this?

Yes, there are a couple partners I know offhand who are interested in testing this.

[Yoav Weiss]

On Wednesday, February 2, 2022 at 5:37:49 PM UTC+1 Dylan Cutler wrote:

Will this be run as a third-party Origin Trial? As a Finch experiment? Other?

This experiment will be run as a 3P Origin Trial, 

So, when the experiment finishes, sites that opted-in to that mode will lose their cookies and their users will e.g. be logged out, etc?
That seems like a deterrent. Is there a way around that? (e.g. migrate the cookies to the default 3P behavior when the experiment is done. Not sure how feasible that is..)

The reasoning behind why we didn't do that is that partitioned cookies allow the existence of multiple cookies with the same host key, name, and path to exist in separate partitions. Rather than coalescing these into one cookie (which one is the right one to keep, after all?), we decided to just remove partitioned cookies from clients' machines when the feature is disabled to provide deterministic behavior.

OK, that makes sense. Given that the breakage at the end of the trial would at worst be equivalent to a cookie clearing, no further concerns on my end.

Happy experimentation!!

 

To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+unsubscribe@chromium.org.

[Daniel Bratell]

Thanks for your answers. I hope it works out fine. (You already have Chris' LGTM so your experiment is ready to go)

/Daniel

[Marcelo Portugal]

Hi,

So I updated Chrome today to version 100.0.4896.60, and now, even when I turn on the partitioned cookie flag, my cookie is not populating the partitionKey as expected. Furthermore, when I was previously playing with it earlier this week on version ~99, although I could see the partition key and behavior working, the partitioned cookies would be blocked once I changed my settings to block third party cookies. So I am unsure on whether I am doing something wrong or if there is a defect with partitioned cookies right now. I was trying with the following url:

https://dim-connect.herokuapp.com/tile

Any help would be appreciated.

Regards,

Marcelo

[Dylan Cutler]

Hey Marcelo,

M100 is when we are starting the origin trial for CHIPS. In order to use partitioned cookies. You must register for the trial, and have your server send the Origin-Trial and Accept-CH: Sec-CH-Partitioned-Cookies response headers to participate in the trial.

If you wish to bypass the OT opt-in mechanism for local testing, you can enable the --partitioned-cookies-bypass-origin-trial flag in addition to the --partitioned-cookies flag. This configuration will allow partitioned cookies from any site regardless of their origin trial status.

All of this is documented in more detail on https://chromium.org/updates/chips.

Best,

Dylan

[Jignesh Malkan]

Hi Dylan,

I am working along with Marcelo on this POC.

We have been able to make progress.

I see 2 options on chrome://flags
partitioned-cookies & partitioned-cookies-bypass-origin-trial

Which flag am I supposed to use?

One thing I have noticed that seemed odd and wanted to see if thats an expected behavior.

We have a small dummy app that just creates a cookie foo-bar and print the value in UI https://dim-connect.herokuapp.com/tile
We embed this in iFrame in our Enterprise App.

From Tab 1 - I go to https://dim-connect.herokuapp.com/tile, a cookie fooBar-1 is created with Partition Key of https://dim-connect.herokuapp.com
From Tab 1 - I log in to Our Enterprise App I see a cookie fooBar-2 is created with Partition Key of http://<enterprise domain name>.com

So far so good.

Now if I go to the Tab 1 and delete cookie fooBar-1, It also seem to delete cookie fooBar-2

Is this expected?

[Dylan Cutler]

Hey Jingesh,

The --partitioned-cookies flag enables the partitioned cookies feature on a Chrome client, however only sites that are participating in the OT can use this feature if only this flag is enabled.

The --partitioned-cookies-bypass-origin-trial flag allows any site to use partitioned cookies regardless of OT participation and is meant for local testing. You still need to use the first flag (--partitioned-cookies) to enable the feature at all.

Upon inspecting the network activity of the page you linked, https://dim-connect.herokuapp.com/tile, I see the problem. If you wish to participate in the OT, then every network response must contain the Origin-Trial and Accept-CH: Sec-CH-Partitioned cookies header. If you do not include these headers, Chrome will revert the cookies to unpartitioned behavior.

Let me know if you have any additional questions,

Dylan