make container aliveness check part of the ipam & dns API

[rade]

...rather than conducting it implicitly when an ID "looks like" a container ID.

Closes #971.

[inercia]

What is the rationale for this? I think that we check the aliveness iff the container id looks like a container id, right?

[rade]

What is the rationale for this?

See the discussion in #971. The purpose is simply to avoid the "looks like a container ID" semantics, which was an implicit part of the API, and instead make it explicitly the callers responsibility to request an ID is treated as a container ID and checked for aliveness.

[inercia]

Maybe "looks like a container ID" is not enough reason for checking for aliveness, but server should check that a valid ID has been provided anyway...

[rade]

server should check that a valid ID has been provided anyway

What does "valid ID" mean? And would the purpose of such a check be?

[inercia]

Your code could happily check with Docker for aliveness of any container name (eg, weave:expose) as long as check-alive is true. I think that kind of checks should be caught earlier and save the query to Docker...

[rade]

So you are saying that if a user makes a mistake, and asked to check the aliveness of something that isn't a container id, then we should catch that and suppress the check with docker? And do what? Add the record? Or not? It's exactly this kind of not-immediately obvious semantics that prompted @squaremo to file the issue.

[squaremo]

Travis seems to have failed for a spurious reason. Restarted that build.

[squaremo]

squaremo removed their assignment just now

@rade If I'm looking into a PR or issue, I will assign myself. If you would like me to look at something, please invite me to do so by @-mentioning me, or by email, or even -- gasp -- by talking to me.