Clarification on the caching mechanism - etld+1 does not seem to share cache

[Liv]

Hi Chromium Security Team,

I’ve been researching cache behavior in Chromium, specifically in the context of shared caches between subdomains of the same eTLD+1.

Based on previous understanding and documentation, it seems that unless explicitly disabled via `Cache-Control`, cacheable resources can be reused across subdomains under the eTLD+1. For instance, if `a.example.com` fetches a cacheable resource, that resource could in theory be reused by `b.example.com` if it's under the same eTLD+1.

However, in testing across multiple providers — such as `amazonaws.com`, `oraclecloud.com` (which is not in the Public Suffix List), and Google/Microsoft-owned domains — I noticed that cache entries are **not** reused across subdomains, even when the resource is publicly cacheable and no restrictive headers are present.

To illustrate this, I conducted a simple test using two storage buckets:

- https://cacher13111.storage.googleapis.com/cached.html  
- https://cacher131112.storage.googleapis.com/cached.html

Both pages attempt to load the same resource (with an almost identical URL in terms of file name, the only difference is the subdomain), and I expected the second to load from cache if the first had already been visited. However, Chromium seems to fetch the resource again from the network.

I’ve also attached a screen recording of this behavior for reference.

Is this the result of Chromium’s cache partitioning mechanisms? If so, I would like to know more about it, or is there something else at play?

I’d appreciate any clarification or documentation you can share on how this is currently handled.

Thanks in advance!

Best regards,  

--

Liv Matan @ Ermetic research team

[thefrog]

awillia@: could you help answer this?

Moving security-dev@ to bcc.