Ad.

Some years ago, Microsoft was working on that. The idea was that you'd submit a query, and when the results of it changed, you'd get a notification. The goal was to have a large system where millions of such outstanding queries could be pending.

What this really means is a database of queries, which is queried when there's a data update to see who needs to be informed.

Anyone know how that Microsoft project came out?

The streaming queries and flash back queries techology is at least 10 years old: https://docs.oracle.com/en/database/oracle/oracle-database/1...

As far as I remember oracle 10 introduced that. In Oracle 11 it actually started working good.

Aurora was developed in 2002 so pushing 20 years (publications at http://cs.brown.edu/research/aurora/).

I used Lucene percolator queries 8 years ago, and it probably wasn't that new at that time.

Anyone thinking about using Confluent as some kind of alternative to a database should read this blog post outlining the myriad correctness problems with ksqlDB: https://scattered-thoughts.net/writing/internal-consistency-...

Am I misreading something here?

> Namely, you can’t write a query whose result updates every time the underlying data change

Yeah, you can. With triggers. Or just a plain old view.

> a query whose result updates every time the underlying data changes

Isn't this exactly what spreadsheet programs like VisiCalc have done since the 1980s?

This is basis for event stream processing systems which nicely supplement database systems. Actually, event stream processing is just inversion of database concept:

In a database you have data to which you direct queries.

In event stream processing you have collection of queries to which you direct data.

Anyone know the price of Confluent's control center product?

Too expensive to consider