kkysen

The latest tech for CBTC uses UWB for train positioning, which has a precision of a few inches.

These pictures show level boarding (minimal vertical gap), they just have horizontal gaps. To the extent that there are some small vertical gaps, that is intrinsic to wheels, which change radius over time as the wheel is worn away. The horizontal gaps should be fixed, though. You'd need gap fillers for them on curved platforms.

Every NYC subway station has had level boarding forever since they were originally built. There can be some horizontal gaps for sure, which they should fix, but it's generally vertically level within the change in diameter of wheels over their lifetime. And if there are gaps, conductors are not helping anyone get on anyways.

The article counts all the off peak and overnight trains, including shuttles. Even out of all of that, only the Franklin Shuttle is 2 cars and will soon be switching to 3 cars.

Only the Franklin Shuttle has 2-car trains, and it'll soon be switched to 3 cars since they're replacing the 75 ft cars with 60 ft cars.

Labor is by far the largest share of transit operating costs.

They already do this on lines with upgraded CBTC signaling. The train runs in ATO (Automatic Train Operation) in which the driver just pushes the start button at each station. So they tried to combine the driver and conductor into one position, but the union fought back and stopped it.

Also, they can't use minimum wage employees for the driver who just pushes a button because the union would throw a fit and might go on strike.

There are other real estate grabs through-running could support. Through-running plans often include a Sunnyside Yards station where Metro-North Penn Station Access (PSA) trains from Hell Gate meet LIRR trains to Penn and Grand Central Madison (GCM), which would also distribute some of the passenger loads out of Penn, which is more constrained. A station here would mean you could also deck over the 180-acre railyard and build skyscrapers (the MTA is exempt from zoning) for some 50k people (about the same population density of Yorkville in the UES) with direct commutes to Penn and Grand Central.

NJT is indeed quite awful with OTP (83%[^1]), train cancellations[^2], reliability[^3], and transparency. However, through-running is also an opportunity for this to improve because, while bad in many ways as well, the MTA is well a class above NJT.

The MTA publishes live GTFS-RT feeds of all of its trains, NYCT, LIRR, and MNRR. There is also https://radar.mta.info/, which shows a map of the entire MTA's commuter railroads' network with every single train, including both MTA passenger trains, Amtrak trains, out of service trains, and work trains. It shows their location, their current speed, their train number, their car length, how late they are, their past and future predicted schedule adherence, and how many people are in each car on the train. The MTA's TrainTime app, which shows much of this information, too, is also leagues better than the NJT app.

So while through-running and integration is an opportunity for NJT to degrade the MTA, it is also a massive opportunity for NJT to improve, especially with regards to data and transparency, which is less of a physical issue. For example, Amtrak's transparency is a lot worse than the MTA's, too, and yet they show up on radar.mta.info all the same, albeit with a bit less info (like passenger counts per car, as their cars don't have automatic passenger counters).

[^1]: The MTA's OTP for LIRR and MNRR are a lot higher (98.5% for MNRR, 95.65% for LIRR). However, these are based on heavily padded timetables, padded by as much as 50% on the New Haven Line (a GCT-Stamford express, Stamford-New Haven local takes about 2:00 today, but could do it in about 1:20 with normal 7% padding and no artificial (non-geometric) speed limits: https://pedestrianobservations.com/2024/02/19/new-york-new-h...). So they essentially add in expected extreme delays so that they never appear late. However, NJT does this, too, and still has much worse OTP.

[^2]: https://www.curbed.com/article/nj-transit-commuter-train-ser... (https://archive.is/smk68)

[^3]: Besides the NEC catenary, which Amtrak owns and is in an awful century-old state, NJT trains themselves have terrible mean distance between failures (MDBF), averaging about 49k miles (https://www.njtransit.com/improve/on-time-performance/rail#:...) compared to about 800k miles for Metro-North's M8s, 16x worse.

The line at Moynihan has appreciably slowed down Amtrak boarding. Pre-Moynihan, Amtrak boarding at Penn was faster.

An extra minute or two can mean a $17 billion expansion, so it matters.

I don't think we do either of those errors. Obviously if the C caller does weird stuff, all bets are off, but we don't mutate through `&T`s or not initialize memory. `MaybeUninit` is only used in a few isolated and carefully checked places. Most of the rest of the buffers are zero initialized, which is usually done for free by the kernel.

This is already done in most places, like browsers. But they want to remove the sandbox for performance reasons.