Docs - 404 link reference

[nmummau]

Steps to reproduce

1. Go here: https://eventuous.dev/docs/infra/esdb/#all-stream-subscription
2. The link in the note takes you to a 404

   Read more about the benefits of using the global event stream.

Expected results

Update docs to point to an active link referencing the benefits of using the global event stream

[linear]

EVE-203

[alexeyzimarev]

That's not easy as I moved my blog to Medium a while ago (don't know if it was the right decision, but anyway). The post is in the archives, and it hasn't been updated for years, although conceptually it is relevant.

I will try to illustrate by this diagram, which illustrated the post:

(the diagram needs to be viewed in light mode)

Basically, I consider things like "stream-based read model" or "aggregate-based read model" a smell. At least when all the read models in a system follow that pattern. Because read models are relatively cheap to build, I advocate for building use-case based read models instead. MyBookings read model in the sample project is a good example. Instead of executing a range query to find all bookings by user id, that read model allows getting the whole thing by id. You don't need to worry about query performance, adding extra indexes, etc. You can even keep it in a fast key-value store (or cheap one, depending on the needs) because all that's required to get the data is the ability to get an object by its primary key.

So, in many cases such read models would consume events from different stream categories. MyBookings, for example, would consume from guest registration (which is not implemented) and prep an empty document. Then, it also consumes RoomBooked and appends entries to a list in the document. But, multi-category subscriptions are very hard to maintain. When instead you consume from the "log" aka $all in KurrentDB terms, you get everything, and it's all in order. That allows you to care a lot less (if at all) about causality, which otherwise would be a requirement.