People keep asking about the dog. His name is Gnarly.

Ask a distributed search engine where a document lives and it consults something that knows: a routing table, a shard map, a coordinator. Take that away and the obvious question is how anything gets found at all.

Every distributed search system built in the last twenty years starts from the same assumption: before you can search data, you must move it into a cluster you control.

That assumption used to be free. It is not free anymore.