| > The announcement itself doesn't go into much detail as to which technologies those would be, though. That's part of the point. They want to be free to compose components depending on their customers needs. CoreOS already includes Fleet (orchestration of individual units), and their are putting effort into Flannel (overlay network), and Kubernetes (orchestration of "pods" of containers). But their market is enterprises that in many cases will want to fit CoreOS into an existing enterprise platform. Nothing in how CoreOS is acting has shown any indication that they think enterprise is "simple". On the contrary, Rocket, while smaller in scope than Docker, if anything is more complex (with the signing, composability of filesets, and implied support for running systemd in the containers). I'm not convinced Rocket is the right thing - they'll have to show me tooling that makes it as simple as Docker first - but some of the decisions appeals to me (a clear spec for the format, that alternative runtimes can conform to; e.g. nothing appears to prevent someone from making a Rocket runtime ("ACE") that uses KVM or Virtualbox or Xen, and hence it could be potentially supported by things like OpenStack, or even shoehorned into a VMware deployment; making things like whether or not an overlay filesystem is used an implementation detail; and while I won't need it often, and it may be an annoyance in some cases, support for signed images). So I just see nothing to imply that CoreOS is a good exponent for some idea of "simple, lightweight enterprise platform". They are providing a set of tools that individually are simple and lightweight, that you can use to start building the lower levels of an enterprise platform with, though, and understandably don't want one of those tools to grow into a platform itself (whether or not Docker will grow into that in a way that doesn't allow the core container execution mechanism to be easily lifted out, is a separate discussion). |
I'd say that's completely valid. I was not trying to imply that the combination of CoreOS and tools such as Fleet, Flannel, Kubernetes etc. claimed to be a "simple, lightweight platform". Indeed, as you point out, the core component, i.e. Rocket itself, may well be more complex than "Docker classic."
I think your comments demonstrate accurately that the overall "Rocket ecosystem" will also address the inevitable enterprise requirements. Since most of the other tools you mention did not appear in the launch announcement, I think it's fair to say that describing Rocket as a response to the emerging 'non-simple Docker Platform' is, at the very least, inviting a comparison that isn't exactly apples to apples.
The overall point I was trying to make goes beyond Rocket, Docker, containers or any specific current technology or tool, however. The way the Rocket announcement was worded simply prompted on ongoing train of thought. Indeed, it's the many conversations in the past months around Docker that I've been involved in that have exemplified the belief/hope/wish that there finally is a new tool that will make boring enterprise problems magically "go away."
Docker, Rocket, and whatever tomorrow's favorite container framework turns out to be, are unlikely to be the last examples of that pipe dream, I think.