In practice, very short texts don't carry very high value so watermarking is (usually) less important. For longer text false positives are not an issue at all since you have a large amount of data to extract your signal from.
I wonder where you got that impression. Several professional watermarking systems for movie studio type content I have worked with (and on) are highly resistant to noise removal while remaining imperceptible.
Based on my research experience and judgment, I have published several top-conference papers in both the detection and diffusion domain, but I haven’t explored the engineering/product side. I believe that if such a system hasn’t been invented yet, it wouldn’t be difficult to create one to remove that watermark using an open-source image/video model and maintain the high quality. Would you be interested in having a further discussion on this?
The wording in the current draft seems to indicate that it applies to "providers of hosting services and providers of publicly available interpersonal communications services". So unless this includes ISPs, I wonder if that means a decentralized P2P service is not covered.
You should scan on the device before it goes on the network. P2P networking or not, the app should include a scanner.
Even if you use an open-source clone without scanner, your contacts most likely will use an app with builtin scanner. Your communications will be scanned on their end.
At that point I'm wondering why we don't also open and scan regular mail at the post office before delivery.
> At that point I'm wondering why we don't also open and scan regular mail at the post office before delivery.
We do open and scan some percentage of regular mail at the post office. It's difficult to find exact sources because the USPS only seems to easily admit to doing it for postage reasons, but it's fairly well known that they search for drugs, bombs, etc. Mail is subject to X-ray scanning and being opened under suspicion of a variety of things happening. When they can't open it themselves, they're also allowed to request permission from the recipient (you can refuse, but then they can go to a judge).
X-Raying import of goods is one thing. What I'm thinking about is more opening letters to read writings and check pictures, without explicit permission nor judge involved.
This is exactly what parent is describing. If you’re on a suspicious person list, or happen to cross paths with one, your mail is likely going to be scanned. I can’t find the source now, but also once read the NSA can intercept a package, modify the contents and send it’s on its way without even a delay in the tracking.
I also wonder how it applies to Matrix, which is encrypted and technically decentralized.
However most users will be using the matrix.org homeserver, which makes it effectively centralized. Though I can still create my own homeserver that talks to matrix.org.
Would matrix.org be forced to offer scanning / a backdoor on the homeserver?
Or would they be forced to add something to the official apps, which is pretty ineffective as there are many client apps.
All in all this proposal seems like a complete mess.
Same here with 20+ years old mail service on the same domain that has never sent spam with correctly configured DNS SPF DKIM DMARC, getting gmail rejections. I noticed a significant improvement after linking the domain to my google account https://support.google.com/mail/answer/9981691
The conclusion of the article says "[...] so far linking (even statically) is done for interoperability, does not [...]". IANAL but it could mean that linking against a GPL library for interoperability is OK in EU, but it does not mean it's true for general case of linking with GPL libraries.
Can’t imagine it’s a big issue; the SSD on my 7.5 year old Skylake MBP is cheerfully claiming 96% lifetime remaining, and seems to be fine. The days of SSDs self-destructing after a couple of years seems to be largely behind us, at least for consumer applications; even low-end stuff has a decent practical lifetime these days.
