shared "groups", feed merging; better relational calculations of GTFS instead of lookup from static into operators into realtime and the reverse.

[kylerchin]

It seems evident that when we combine realtime and static data, they should be stored together as a group rather than this jungle. Perhaps feeds are still useful for downloading. Right now, have to intepret this on the frontend. When we do trip arrivals, we need to store the data together into some sort of association.

I've made a few diagrams. Hexagons are the operators. Squares are the static feeds (zip files downloaded occassionally), rounded are realtime feeds (protobuf or something similar).

We need to stop doing this lookup pattern through operator and just group everything together.

The problem is that transit.land is so useful and it will take a few hours to manually assign each agency into a group if we do it manually. And I want to prevent collisions. We also have to give them consistant network ids. Maybe we can do this via a web interface?

Transit app uses the idea of single networks as agencies.

New York City MTA

Note I didn't even get into the insanity that is MNR and LIRR.

Germany

f-germany~regional~rail Apparently has 103 operators under it but goes back to a single GTFS realtime. Theoretically, we can throw away the transitland defined operators in this case and just use the single file.

United Kingdom

f-busdftgov~uk is basically the same deal as Germany.

Amtrak

CC is inside the San Francisco Metropolitan Transportation Commission's static feed and realtime feed, so we should probably delete the data inside amtrak feed and transfer the data over to the Amtrak data system.

Most American agencies

1 to 1 correlation. Static feed <-> Operator <-> Realtime feed

Yokohama, LA Metro, Washington DC, etc

Both are like this. This applies to Washington DC and other cities that have split feeds.

f-subway-static <-> agency-operator-id <-> f-subway~rt
f-bus-static <-> agency-operator-id <-> f-bus~rt

Do we split? or merge?

Conclusion

My brain hurts

[kylerchin]

According to https://backend.catenarymaps.org/getinitdata there are currently 1411 static feeds, 2277 operators, 374 realtime feeds.
1785 total feeds.
So if we spend ~1 min thinking about each feed, that's 29.75 hours of manual work.