Visual Studio Code is designed to fracture

Nevermind the face-value argument of performance and usage data turns into "give me a debuglog that includes every program you have open, a whole bunch of personal computing hardware information which can be trivially tied to PII".

The argument that telemetry and server-side knowledge is safe entirely depends on reducing the argument down to an idealized metaphor that doesn't exist in practical reality, and then rejecting all forms of evidence that don't fit the mould.

Unfortunately in this scenario, telling the truth about what needs to be kept private, is in itself, a clue to what is being kept private. So no communication can happen back up the chain if the client wants to keep something private. When the server becomes untrustworthy, so must the client.

> telemetry is just another way to die by metadata.

If your threat model is a cruise missile from the military, then I actually agree - don’t use VSCode. For the rest of us who don’t live in fear of being murdered by a violent government at any moment, it’s a reasonable trade-off.

I don’t see how I could be de-anonymised by these logs but I’m happy to be proven wrong. If you could cite which metrics are being collected and how you would de-anonymise someone with that data, that would be great.

We disagree on this, so I'll take that into account every time I see one of your comments on this site. (Not really, but you get my point.)

VSCode could collect anything at all, but let's be hypothetical, yet specific so you understand the process. I'm going to use these two pages as reference [0]This RoboLeary blog post describes what VSCode collects (more than just performance, but we'll leave that for now) and [1]This Github repo shows an example log file of just the telemetry data you can voluntarily opt out of.

[0]https://www.roboleary.net/tools/2022/04/20/vscode-telemetry....

[1]https://gist.github.com/robole/97f6e1c4dc888ae0a49a7683e5494...

In the github example log file, there's 9k lines of code, with thousands of xml style entries with properties. Those properties can be any value you want for the purposes of de-anom. It doesn't matter what it is.

Your database back at Microsoft HQ (or wherever this stuff goes) will have a huge ocean of telemetry logs, so how do we find out which one is yours? See below.

In those thousands-of-entries, some of them will differ statistically from the average of all log reports. It could be the time it takes for a menu to load in milliseconds, it could be the filepath of an extension, it could be a specific hardware report, it could be a collection of exceptions that keep throwing. It doesn't matter what it is, it matters that it differs from everyone else.

Some of those reports that differ from the herd (statistical average) can be collected together to make a fingerprint of an individual machine. Maybe your machine always takes slightly longer than average to open a menu, or has a different extension set, whatever. The tiniest difference is still a difference.

This fingerprint of your machine can be correlated statistically to be likely the same machine again and again across telemetry logs, across time. Your fingerprint keeps showing up across multiple logs, across multiple days, we know a time of computer usage. The fingerprint does not need to be perfectly separated from every other fingerprint in the bell curve, but we can solve that later on. As long as it's narrow enough (only so many conflicting fingerprint reports) we can narrow it down with added metadata.

How do we know who is running VSCode and doing this work on this fingerprinted machine? All the other 'harmless' metadata like IP address, VS licenses, Microsoft accounts, linked email addresses, and most importantly all the data harvesting Microsoft will not let you turn off.

So this isn't a problem right? Who cares if Microsoft can trivially (and likely automatically) de-anonymize me. No junior engineers have access to the data, only Edward Snowden et al. can see it, so who cares?.

Management can easily take access to the database of fingerprints and build analysis tools and programs the end user and the junior engineer alike, cannot see or understand what decisions are being made from it. So if you and I (assuming we're both grunts in software) can implement a perfect telemetry system that collects "no useful category of information" and the database is kept from public eyes, we can both still lose out on what management does with that information, as there is more to be revealed in it than we want.

So if military is happy to kill people based on metadata and our advertising data is being sold left-right and centre and insurance and health data is going haywire, what assurance does any computer user on a windows machine, or VSCode user have, that their data is not going to be used to make decisions arbitrarily (from the end user perspective) against their will?

If I don't want VSCode to know that I'm working on a project at 3am, that's my business and revealing what Harm that causes and why I might want to keep that private, reveals a clue about what is being kept private, and thus cannot be shared. Steve Jobs would have completely removed the collection of meta-data from secret product development teams and no-one would have batted an eye. When a random joe does it, there's all these questions about what Harm does it do, as if someone had no reason to keep a secret after Steve died.

If you want transparency from the rank-and-file, get it from what decisions management is making with metadata too and tell me what Harm Does It Cause Microsoft Management To Reveal Their Truth. Heaven forbid I suggest reality is a two-way street.