CouchDB scaling
The CouchDB is appealing for those who want to expose their datamodel to the world via web-services (REST). RWW ponders whether when coupled with Javascript-y clients, could this model of development cut out the server-side presentation layer? We expose a lot of data via Web Services, but strangely we are rarely exposing raw data models. There is often a layer of indirection, translation, validation and BP processing associated with these services, rather than a thin-veneer. That could be just a style thing on our part.
Having read about CouchDB, I could see the appeal, certainly for internal projects, but it seemed a little bit alpha-ish for my purposes. Then we ran up against it in anger, when using Chef for ops automation. Basically it can began to fall over under heavy write load - a lot. Much tweaking, auto-compacting and other general pottering about with bleeding edge code and it is functional, but not ideal. Then I came across the blogpost from Jonathan Ellis's - CouchDB: not drinking the kool-aid.

Having read about CouchDB, I could see the appeal, certainly for internal projects, but it seemed a little bit alpha-ish for my purposes. Then we ran up against it in anger, when using Chef for ops automation. Basically it can began to fall over under heavy write load - a lot. Much tweaking, auto-compacting and other general pottering about with bleeding edge code and it is functional, but not ideal. Then I came across the blogpost from Jonathan Ellis's - CouchDB: not drinking the kool-aid.



