What is the minimal set of client hints needed for the best analysis result?

[istreeter]

Background: we have implemented yauaa into Snowplow Enrich (see snowplow/enrich#639). And it's working great.

In testing the feature, I get a much better yauaa result (meaning more information) if I configure a site to send the Sec-CH-UA-Full-Version-List header. However, if I configure the site to add any extra header (say Sec-CH-UA-Arch or Sec-CH-UA-Bitness) then it does not seem to improve the result any further.

Because of this, in the Snowplow documentation we are currently recommending to just send the Sec-CH-UA-Full-Version-List header. I could not see any reason to document adding the other headers.

So my question is... does this observation match with how you expect the analyzer to work? Or should I expect that adding extra headers like Sec-CH-UA-Arch or Sec-CH-UA-Bitness will give a different analysis result?

I know that client hints is an evolving browser feature -- so is it likely that in future we will need to request extra hints to still get the best result?

Thank you for your great work maintaining this very helpful project.

[nielsbasjes]

The current situation is that the useragent still has some valuable information which is now duplicated in the client hints.
With browser builders stripping the information from the useragent and moving the info to the client hints you can see the need for the client hint increasing over time in the near future.

So right now you'll see that the client hints in many cases do not have added value because both the UA and CH contain the same information.

You mention the Sec-CH-UA-Arch and Sec-CH-UA-Bitness which currently are the only way to detect if a Mac is running an M1 (ARM based) cpu. Without these hints this testcase would actually indicate an Intel cpu instead of ARM:

yauaa/analyzer/src/main/resources/UserAgents/ClientHints-Tests.yaml

Line 913 in ecf9ee2

DeviceCpu : 'ARM'

So right now (Q1 2023) it is really only some of the Chromium based browsers (Chrome, Edge but not Brave) that do the client hints.

The code that does this augmentation is here https://github.com/nielsbasjes/yauaa/blob/main/analyzer/src/main/java/nl/basjes/parse/useragent/clienthints/ClientHintsAnalyzer.java

And many of the testcases are here:
https://github.com/nielsbasjes/yauaa/blob/main/analyzer/src/main/resources/UserAgents/ClientHints-Tests.yaml

So right now my very simple advise is to request all of these client hints and persist them.

If you feel like finding out what the impact is of removing a client hint you can do perhaps something like download the code and comment the header you want to see the impact of and then run the tests to see what the failure impact is.

[nielsbasjes]

A bit deeper dive into the fields and their "added value" and why to keep them or not.

Client hint	Example	Keep it?	Why
Sec-Ch-Ua	" Not A;Brand";v=“99”, “Chromium”;v=“100”, “Google Chrome”;v=“100”	Yes	You may not have the Sec-Ch-Ua-Full-Version-List with the exact versions.
Sec-Ch-Ua-Full-Version-List	" Not A;Brand";v=“99.0.0.0”, “Chromium”;v=“100.0.4896.75”, “Google Chrome”;v=“100.0.4896.75”	Yes	Is the better variant of Sec-Ch-Ua but it may not be present.
Sec-CH-Ua-Full-Version	“100.0.4896.75”	No	This field is deprecated in the standard. Also this info is also present in the Sec-Ch-Ua-Full-Version-List
Sec-Ch-Ua-Mobile	?0	Yes	In the (very rare) case where we cannot determine if it is a phone or tablet this flag determines the end result.
Sec-Ch-Ua-Platform	“Windows”	Yes	Needed in the very common case of bad version info in the useragent.
Sec-Ch-Ua-Platform-Version	“0.1.0”	Yes	Needed in the very common case of bad version info in the useragent. This "0.1.0" means "Windows 7" because the Platform says "Windows".
Sec-Ch-Ua-Arch	“x86”	Yes	The only way to determine a MacOS system is running an M1/M2 (ARM) instead of an Intel CPU
Sec-Ch-Ua-Bitness	“64”	Yes	Often not present in the useragent
Sec-Ch-Ua-Model	""	Yes	Often not present in the useragent (brand and device info).
Sec-Ch-Ua-Wow64	?0	No	The only thing this says is that this is Windows (use Platform) and that it is 32 bit software running on a 64 bit system.

To simplify it all I would

• Ask for all of them (regardless if I say No above) because the browsers are currently in flux in what they put in these fields.
• Persist all headers starting with Sec-Ch-Ua along with the User-Agent

[istreeter]

Thank you this is an extremely helpful answer!

[dletendre-hc]

Thank you Niels!

[nielsbasjes]

FYI: I've added this to the documentation https://yauaa.basjes.nl/using/clienthints/#which-client-hints-do-i-really-need