Hacker News Reader

Comment by mping | Hacker News Reader

The immediate application of it is that when you choose a DB, you should be able to tell if it's either CP or AP. The field is full of subtleties (eg Network is not 100% reliable so no system should be CA - it's way subtler but that is the gist of it. I'm no expert btw).

Lots of DB vendors have tried to circumvent the therorem by introducing their own concepts or stretching the definitions. "We can beat the CAP theorem"

View on HN

is CA even possible in a distributed system though? I mean, it’s not like “the network never fails” is a design decision you’re allowed to make (short of removing the network altogether and therefore not being a distributed system at all)

That's what Google spanner is claiming, that Spanner is Consistent and highly-Available (nine fives, or 99.99999% of the time available), which "in practice" they say is "CA".

From the perspective of the CAP theorem, Spanner is unquestionably a CP system, as the original paper [1] makes clear. (It relies on Paxos to elect leaders, and Paxos chooses consistency over availability.)

The "CA in practice" claims are purely marketing.

[1]: https://storage.googleapis.com/pub-tools-public-publication-...

It would be nice to have an overview of existing/possible workarounds.

The point is that eventually it's found out that there aren't any. They're all snake oil.

Usually companies find this out the hard way, with catastrophic loss of business data.

Reply on news.ycombinator.com