cluster: rename worker.destroy() to worker.kill() again

[bnoordhuis]

It was renamed from kill() to destroy() in 5f08c3c for no apparent reason AFAICT.

To keep the mental disconnect with the child_process module as small as possible I want to name it (or alias as) kill() again.

/cc @AndreasMadsen

[AndreasMadsen]

There is a couple of reasons however nothing solid.

• The worker.destory() don't take a signal child.kill() dose.
• When dealing with workers it is quite important to indicate that this is a bad "shutdown" (to use a 3th work), .destroy() indicates that. It translate to the socket.destory().

In general the main decision I made in the cluster refactor was about "should this match child_process.fork or the net.Server API" since a cluster worker is both (it has a listening and exit event).
I chose the net.Server as the primary base, since the v0.6.x cluster module was build on child_process.fork and did clearly prevent a lot of extensibility to be added in the refactor. But to maintain some backward compatibility there would be a .process event, linking the server like object to a process.

It funny I remember having this discussion before, perhaps on IRC or in my head, who knows?

[isaacs]

@bnoordhuis @AndreasMadsen I have very little opinion on this, except that it'd be nice to not break backwards compatibility with 0.8. Maybe an alias?

[AndreasMadsen]

If it should be renamed then do the following:

• Alias .destroy() to .kill()
• Deprecate .destroy()
• Add optional signal argument to .kill()

Especially the signal argument seams important, since it will then look like the require('child_process').fork().kill method.