The problem with retractions should be easier to deal with in the online domain than in print. Make it a rule that whenever an article which is subject to a retraction is requested that the retraction is automatically, and prominently, included in the page served.
If not to the ISP, who should Netflix or YouTube pay the costs to? They don't bear them directly, because the traffic is necessarily much cheaper for them to carry than it is for the ISP.
They should pay the costs for their connection to/at the Internet Exchange Points (IXP) the same way the ISPs pay for their connections to the IXP.
Nor carry anything which is mains driven rather than battery operated - so no corded only electric shavers in your hand luggage (Do long haul flights have a shaver socket in the bathrooms?)
Why would the takedown notice be sent to the ISP? Would it not be sent to one of the contacts in the WHOIS for the domain?
Do most DMCA notices not just require taking down a single URL rather than all pages in a domain? In which case it is the site owner, or their webmaster, who has to action the notice, so sending the notice to the ISP will just delay the process.
I disagree with your disagreement of the analogy. What if your house is rented or you have put the physical letter is in a safety deposit box in the bank? In both these cases the physical location is owned by someone else and you are just renting the space. Is this any different from you renting the e-mailbox space on the google (or other ISP) servers?
Someone should tell Goldman Sachs that you cannot unsend an email. Usenet articles can be cancelled, even though most servers ignore cancels, but like snail mail, once email is posted it cannot be recalled.
But having opened the safe, can they force you to 'decode' the entries on a paper document which are written in a code or cipher? If not, then they should not be able to force you to decrypt an electronic document which is written in 'code'.
Even before AOL there was CompuServe with its octal user IDs
In which case the manufacturer should have upgraded OpenSSL and released (maybe even pushed) the update.
When the Heartbleed exploit was announced, all users of vulnerable openssl versions should have upgraded.
In which case talking about it should never be allowed as refusing to either confirm or deny when previously having denied would strongly indicate that a gagging order had been issued.
Or even worse when you do not even leave your home country but your phone happens to connect to a mast in a neighbouring country.
Or to learn the techniques used by the Faceless Men in the Song of Ice and Fire series.
Hiding the destination need not be difficult. You just do the electronic equivalent of putting a coded small ad in a newspaper. Everyone can read it, but only the intemded recipient can decode it and there is no indication as to whom the message is intended for.
Even that does not help. These mechanisms only authenticate you to the bank, not the bank to you. A spoof bank site could still request the OTP password or output from the dongle and accepr whatever response you give.