Description
When building and pushing docker images in a shared environment (e.g. on a jenkins worker) there is always a possibility that between a build command and the following push command that someone else pulls the same tag as the one that is to be pushed.

This can result in a push of the other image instead of the one intended. The reason for this issue is that we can't set a source image when pushing, we have to use the same name on the local docker host as the name in the registry.

Steps to reproduce the issue:

1. User 1: docker build -t myregistry/an-image:stable .
2. User 2: docker pull myregistry/an-image:stable
3. User 1: docker push myregistry/an-image:stable <- this will push the old image and not the one User 1 just built.

Suggestion

Add an option to docker push that allows users to specify which source image they want to push.
Example: docker push --source <image-id-or-name> NAME[:TAG]

This addition would allow users to minimize the risk of pushing the wrong image to the registry.
Adding an option ensures backward compatibility of the command.