I've read that even though T-Mobile Home Internet offers IPv6, it blocks all inbound TCP connections even on IPv6. Its gateway appliance offers no port forwarding nor DMZ.

Once there are more Internet subscribers than IPv4 addresses, then by the pigeonhole principle, some subscribers aren't going to have the sort of dedicated IPv4 address needed to accept a TCP connection. Therefore every ISP is shit.

The problem with IPv4 is that there aren't enough possible network addresses for an ISP not to be shit.

A reverse-SSH tunnel requires one of two things: either your local computer is on a network that can accept inbound connections, or there's a relay ($) in the middle accepting connections from both the client and the server.

I believe there's a substantial qualitative difference between "extra effort" and "cost", especially when the latter is a recurring cost payable to the rent-seekers that run relays.

That's the easy part. Getting your ISP to forward a port to your LAN is the hard part.

Is it a good thing that everyone who needs to connect to a home NAS or remote desktop from outside the home LAN be required to subscribe to a relay like Pinggy, Tailscale, or Hamachi, on top of what the user already pays the ISP per year for an Internet connection?

When an entire neighborhood shares an IPv4 address through ISP-controlled carrier-grade NAT, how does a device on subscriber premises receive an incoming TCP connection? How would the NAT appliance even know for which subscriber's device the connection is intended?

Consider a subscriber whose home LAN is behind the ISP's carrier-grade NAT, and the subscriber wants to connect to a home NAS or remote desktop from outside the home LAN. Other people have recommended that such a subscriber additionally subscribe to a relay service such as Tailscale or Hamachi. And if they want visitors from the Internet to reach their home server, it gets even more expensive.

What makes IPv6 difficult for a lot of people, such as myself, is that we're in no position to test on it. Last I checked, US fiber ISP Frontier Communications still refused to deploy IPv6 in my city, and the alternative was Comcast.

Lean into that. Make an episode where poodle!Betty tries to fit into a human-dominated society but fails to pass as human, with people calling her the B word, and have it end on a cliffhanger as she is prepared for plastic surgery.

Names of characters used in more than one work or in merchandise are generally subject to trademark, not copyright. The US Supreme Court has refused to enforce the Lanham Act, which relates to trademarks and passing off, in such a way that it would extend the effective term of an expired patent or copyright. Kellogg Co. v. National Biscuit Co., 305 U.S. 111 (1938); Dastar Corp. v. Twentieth Century Fox Film Corp., 539 U.S. 23 (2003).

The text of Title 17, United States Code, section 305:

All terms of copyright provided by sections 302 through 304 run to the end of the calendar year in which they would otherwise expire.

From House Report 94-1476, linked from "Notes" on the same page:

Under section 305, which has its counterpart in the laws of most foreign countries, the term of copyright protection for a work extends through December 31 of the year in which the term would otherwise have expired. This will make the duration of copyright much easier to compute, since it will be enough to determine the year, rather than the exact date, of the event from which the term is based.

How would a congestion charge work in cities that run one bus an hour on each route and don't run any buses at night or on Sundays? (Source)

When the seller requires payment before providing a product or service, there's no debt and therefore no opportunity for "legal tender" to apply.

The desktop PC industry has had to deal with customers bringing their own DRAM DIMMs and other components across an upgrade for decades. Enthusiasts' desktop PCs are such a ship of Theseus that Microsoft had to choose one component (the logic board) as the seat of license for OEM Windows. Laptops, by contrast, have tended to use fewer standardized, user-replaceable components to save the bulk and weight of connectors. SODIMMs have historically been one of the few components that users can take with them, but the switch to DDR5 CAMMs in many newer laptops has complicated this.

Regarding putting DRAM modules lower than on-package DRAM in the memory hierarchy:

When DDR5 was pushing the bandwidth limits of the SODIMM form factor, Dell created CAMM and had it standardized through JEDEC. I'm guessing that the shift from SODIMM to CAMM staved off the need to demote DRAM modules by a level in the hierarchy. Maybe some design decision in the Apple M series requires a much faster RAM interconnect then Intel or AMD.

Apple processors in the M series have DRAM on package, which provides better signal integrity than crossing a logic board. Apple could have added a SODIMM or CAMM slot for additional RAM to be used as a RAM disk holding a swap file. But they chose not to, instead putting swap in the SLC intake buffer of the soldered-in SSD.

I agree that it is a silly idea. What ought to be the workaround for the phone operating system publisher taking 30 percent of all in-app purchases and the record label taking the now-standard 70 percent, leaving zero percent for hosting the download store?