Intent to Extend Experiment: Compression Dictionary Transport

[Tsuyoshi Horo]

Contact emails

ho...@chromium.org, pme...@chromium.org, yoav...@chromium.org, kenji...@chromium.org, deno...@chromium.org

Explainer

https://github.com/WICG/compression-dictionary-transport

Specification

https://datatracker.ietf.org/doc/draft-ietf-httpbis-compression-dictionary/

Design docs

https://docs.google.com/document/d/1IcRHLv-e9boECgPA5J4t8NDv9FPHDGgn0C12kfBgANg/edit

https://github.com/WICG/compression-dictionary-transport

https://datatracker.ietf.org/doc/draft-ietf-httpbis-compression-dictionary/

Summary

We are running the second round of the Origin Trial until Chrome 125. The design of the feature has also evolved during the origin trial and RFC process. We’d like to run a third round of the Origin Trial to get more feedback on the updated design.

Blink component

Blink>Network

TAG review

https://github.com/w3ctag/design-reviews/issues/877

TAG review status

Closed

Risks

Interoperability and Compatibility

Interoperability and Compatibility risk are low. This feature introduces a new compression method for transporting resources over HTTP. Web sites can know the browser support for the new feature by checking `document.createElement('link').relList.supports('compression-dictionary')`.  The feature is a negotiation between servers and clients that involves a server specifying that a resource should be used as a dictionary for future requests with a ‘Use-As-Dictionary’ response header. Clients that don’t support the feature will ignore the header and nothing further will happen.

This feature is an opt-in feature. And the dictionary storage is isolated using the top level site and the frame origin as the key. That means, if there is no dictionary registered for the site, the behavior of Chrome will not change while browsing the site. Also this feature is only usable within a secure-context so this feature will not increase the risk of having network proxies meddle with the content’s encoding. For enterprises that have deployed HTTPS-intercepting proxies that do not properly handle unknown encodings there is an enterprise policy exposed to disable the feature. The feature is currently enabled only over connections using a well-known trust root for the secure connection which eliminates the risk of security devices and proxies breaking traffic. The roll-out plan for production has steps to remove the trust root restriction and enable support in enterprise environments where intercepting proxies are common.

Gecko: Positive (https://github.com/mozilla/standards-positions/issues/771)

WebKit: Positive (https://github.com/WebKit/standards-positions/issues/160)

Web developers: Positive

Other signals:

Ergonomics

To reduce memory usage in network services, dictionary metadata is stored in a database on disk. And to avoid performance degradation for normal requests that do not use a dictionary, the reading of this metadata is designed not to block network requests. In other words, if the reading of metadata from the database is not completed before the request header is ready to be sent to the server, the dictionary may not be used even if it is already registered in the database.

Activation

To adopt this feature, web developers need to make changes in their web servers or build processes for static resources. Currently there is no major server software which directly supports compression dictionaries. Some CDNs have shared interest in supporting shared dictionary compression (e.g. publicly mentioned in a blog post by Cloudflare).

Security

Chrome registers the response as a dictionary only when the response is CORS-readable from the document origin. Also we use a registered dictionary to decompress the response only when the response is CORS-readable from the document origin. Additionally, the dictionary and the compressed resource are required to be from the same origin as each other. So this should not introduce any new attack vector of information leaks.

The dictionaries are partitioned with the storage cache and are cleared whenever cookies or cache is cleared to ensure that the dictionaries can not be abused as a tracking vector.

WebView application risks

Does this intent deprecate or change behavior of existing APIs, such that it has potentially high risk for Android WebView-based applications?

No

Goals for experimentation

We would like to collect feedback on the updated API design of Compression Dictionary Transport feature. Also, we would like to continue some experiments using this feature to measure its performance impact.

The difference from the previous Origin Trial:

- Renamed the link rel attribute for fetching the dictionary from "dictionary" to "compression-dictionary".

- Using "dcb" and "dcz" content encoding in place of "br-d" and "zstd-d". The new encodings incorporate the dictionary hash.

- Removed the dictionary hash check logic using the "Content-Dictionary" response header, and instead checking the hash within the encoded response body.

Ongoing technical constraints

None

Debuggability

We have introduced chrome://net-internals/#sharedDictionary. Using it, web developers can manage the registered dictionaries. Also web developers can check the related HTTP request and response headers (Use-As-Dictionary, Available-Dictionary, Accept-Encoding, Content-Encoding).

Any errors in processing responses as a result of dictionary compression issues are reported as issues in Dev Tools. This includes things like dictionary hash mismatches and cors-readability failures.

Will this feature be supported on all six Blink platforms (Windows, Mac, Linux, Chrome OS, Android, and Android WebView)?

Yes

Is this feature fully tested by web-platform-tests?

Yes. https://wpt.fyi/results/fetch/compression-dictionary

Flag name on chrome://flags

chrome://flags/#enable-compression-dictionary-transport-backend chrome://flags/#enable-compression-dictionary-transport

Finch feature name

CompressionDictionaryTransportBackend CompressionDictionaryTransport

Requires code in //chrome?

True

Tracking bug

https://crbug.com/1413922

Launch bug

https://launch.corp.google.com/launch/4266286

Estimated milestones

OriginTrial desktop last	129
OriginTrial desktop first	127

OriginTrial Android last	129
OriginTrial Android first	127

Link to entry on the Chrome Platform Status

https://chromestatus.com/feature/5124977788977152

Links to previous Intent discussions

Intent to prototype: https://groups.google.com/a/chromium.org/g/blink-dev/c/-qYpLo9DTjw/m/JX6kbUOtBQAJ

Intent to experiment: https://groups.google.com/a/chromium.org/g/blink-dev/c/NgH-BeYO72E/m/oup5DpbxAAAJ

Intent to extend Origin Trial: https://groups.google.com/a/chromium.org/g/blink-dev/c/7YdP2YxRQn4/m/9oPxLDdjAAAJ

[Mike Taylor]

Hi there,

Would you mind commenting on progress for the following items, per OT renewal guidelines:

Draft spec (early draft is ok, but must be spec-like and associated with the appropriate standardization venue, or WICG)
TAG review (see exceptions)
bit.ly/blink-signals requests
Outreach for feedback from the spec community
WPT tests

thanks,
Mike

--
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/CADk0S-UpM8k_NGuqbuA%2BNuC%2BTdGDjGzUt0Fp05Y8pHROjH-WZA%40mail.gmail.com.

[Yoav Weiss (@Shopify)]

On Wed, May 29, 2024 at 4:31 PM Mike Taylor <mike...@chromium.org> wrote:

Hi there,

Would you mind commenting on progress for the following items, per OT renewal guidelines:

<API owner hat off / feature champion hat on> 

Draft spec (early draft is ok, but must be spec-like and associated with the appropriate standardization venue, or WICG)

Recently published with above-mentioned changes. 

+Patrick Meenan can probably add precision there, but it's making good progress in the HTTPWG.  

TAG review (see exceptions)

https://github.com/w3ctag/design-reviews/issues/877 

bit.ly/blink-signals requests

Both Mozilla and WebKit are positive (with ongoing discussion about some details with Mozilla folks).

Outreach for feedback from the spec community

Regular HTTPWG and WebPerfWG engagement. 

WPT tests

https://wpt.fyi/results/fetch/compression-dictionary?label=experimental&label=master&aligned 

[Patrick Meenan]

On the spec side of things, there is one more issue outstanding in the IETF discussion but it's not API-impacting and we expect that this latest draft, which this OT implements, has the final API shape. There will be some tweaks around the edges as we go through the final steps of the RFC process but the V3 OT will give sites something to test against that matches what we expect to ship.

There are some fairly substantial changes from the previous OT that it would be useful to get feedback on. In particular, the change to the file format that embeds the dictionary hash into the file itself instead of being a separate response header.

[Mike Taylor]

Thanks all. LGTM to extend from 127 to 129 inclusive.

[Panos Astithas]

Hi Tsuyoshi,

Is this a request to extend the V1 OT as it appears in Chromestatus and in the title of this thread or are you trying to create a new V3 trial as it seems to be the intent based on the content of your intent? Note that V1 ended at M125, so teh extension would be for 4 milestones.

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/f9f7a485-94d3-45cf-a0b1-3021c6d7526a%40chromium.org.

[Patrick Meenan]

Sorry, probably some confusion with the process.

This is for the 3rd round of OT on the platform status entry (CompressionDictionaryTransportV3) so "extend" may not be the right terminology.  V1 really ended at 122 and we had the same confusion the last time around and the V2 trial was created that went from 123-125 (and is the current active trial).

I'll leave it to Tsuyoshi so I don't accidentally break anything, but I assume we need to mark the v3 trial as the active stage.

[Tsuyoshi Horo]

Ah, sorry for the confusion.

This request is for the V3 Origin Trial.

V1 OT was from 117 to 122.

V2 OT was from 122 to 125.

And this V3 OT is from 127 to 129.

[Jason Robbins]

Mike or other API Owners, still approved given that this is actually requesting a new OT?

Thanks,

jason!

[Domenic Denicola]

LGTM for a new OT from 127 to 129.

(Speaking generally, I think starting new OTs is better than extending existing ones, so I am glad the team has taken this route. From an ecosystem perspective, new OTs make it easier for the team to make breaking changes, and encourages OT participants to continually re-engage with the process.)

[Mike Taylor]

Thanks Domenic - I agree.

[Jason Robbins]

Tsuyoshi,

Previously your third OT stage on ChromeStatus got associated with your second trial on OT Console.  That is probably due to a bug in ChromeStatus. We'll look into that.

To let you continue with requesting this new trial, I have cleared out the trial ID for that third OT stage on ChromeStatus.  Now you should see a "Request Trial Creation" button on that stage that looks like:

https://screenshot.googleplex.com/4RZcpmCHJgxSnaT

You will need to change some fields in the form to indicate that this is for your "V3" trial.

Please note that each distinct trial requires a distinct trial name with an entry in 

https://chromium.googlesource.com/chromium/src/+/main/third_party/blink/renderer/platform/runtime_enabled_features.json5

The "Request Trial Creation" form handler now checks that file when you submit, so you may need to land a change to that file before you can successfully submit the form.

Thanks,

jason!

[Tsuyoshi Horo]

Hi.

On Sat, Jun 1, 2024 at 3:04 AM Jason Robbins <jrob...@google.com> wrote:

Tsuyoshi,

Previously your third OT stage on ChromeStatus got associated with your second trial on OT Console.  That is probably due to a bug in ChromeStatus. We'll look into that.

To let you continue with requesting this new trial, I have cleared out the trial ID for that third OT stage on ChromeStatus.  Now you should see a "Request Trial Creation" button on that stage that looks like:

https://screenshot.googleplex.com/4RZcpmCHJgxSnaT

Sorry, I can't find the "Request Trial Creation" button.

I still see the "Request a trial extension" button, not the "Request Trial Creation".

https://screenshot.googleplex.com/9PNnFQLkQBTVwdU

And the "View Origin Trial" button is linked to the previous V2 Origin trial console URL.

[Jason Robbins]

Hmm, after I cleared out the wrong data, something filled it in again.  I will try to track down that ChromeStatus bug today.

Thanks,

jason!

[Jason Robbins]

OK, I think that I found the problem and fixed it in ChromeStatus database.  I reran the cron job that previously overwrote my simpler fix and it did not overwrite my more-complete fix.  So, you should now see the "Request Trial Creation" button on the V3 OT stage, and it should still be there until you press it and submit the form.

Thanks,

jason!

[Tsuyoshi Horo]

Thanks!

I just submitted the creation request.

[Tsuyoshi Horo]

Hi, blink-dev@ folks.

Let me announce the start of the third round of the Origin Trial for the Compression Dictionary Transport feature.

If you are already experimenting with this feature and you want to continue the experiment after Chrome 127, you'll need to register for a new token (CompressionDictionaryTransportV3).

There were changes since the first trial. So please check this list of changes if you have any trouble while experimenting with the Compression Dictionary Transport feature in Chrome 127. 

Thank you.