Transparent HTTP caching as a way to avoid leaking metadata is not pro-privacy. It only works because the network is always listening, to both metadata and message content. The reason why people worry about metadata is because it's a way to circumvent encryption (and the law). Metadata is holographic[0] to message content, so you need to protect it with the same zeal content is protected.
But letting everyone have the message content so that metadata doesn't leak isn't helpful. Maybe in the context it was deployed, where pervasive deep packet inspection was only something China wasted their CPU cycles on, your proxy made sense. But it doesn't make sense today.
[0] X is holographic to Y when the contents of X can be used to completely reconstruct Y.
How it metadata holographic? Sure, you can know when I communicated to a particular individual, and even the format and size of the message, but it doesn't include the exact message, right?
Gordon Welchman first productionized “traffic analysis” in WW2 at Bletchley Park.
When in his retirement he tried to write about it, it was his work on traffic analysis more than his disclosing that the allies had cracked enigma that most worried the NSA who tried to stop him publishing.
Traffic analysis is in many ways more valuable than the contents of the messages.
I won't say that metadata isn't valuable, but I still don't think it's holographic. You can tell I WhatsApp my friend every day around noon, so we're probably talking about lunch, but you don't know that today I had a tuna sandwich.
Old thread but I think there’s a wood and trees thing here.
Traffic analysis is king because who you communicate with is a low noise signal and what you communicate is usually noise.
This is well known for police work and military intelligence etc.
It’s also true for ad sales. Ad networks want the trackers on sites so they can build up a profile of you based on metadata not the content of the pages you visit themselves.
Yeah, that's all fine, but the original claim was that:
> Metadata is holographic[0] to message content
...
> [0] X is holographic to Y when the contents of X can be used to completely reconstruct Y
To say something is holographic is a claim about data, not of value. I totally buy that metadata is valuable, it could even be more valuable than the contents, but it's not a means to reproduce the message content. My ISP can tell where I bank, which is certainly valuable to observers, but it can't tell my password or the contents of my accounts, all of which I transmit. That's not a holographic reconstruction.
But letting everyone have the message content so that metadata doesn't leak isn't helpful. Maybe in the context it was deployed, where pervasive deep packet inspection was only something China wasted their CPU cycles on, your proxy made sense. But it doesn't make sense today.
[0] X is holographic to Y when the contents of X can be used to completely reconstruct Y.