Validation for routable IP in HostName

[bherw]

It would be helpful to have a HostNameParameters option to reject HostNames that cannot be routed to.

For example, 0.0.0.0 is a valid IP address, but I'm trying to validate configurations that I need to push down to other systems to get them to connect to those hosts, and 0.0.0.0 isn't going to be a valid IP to connect to.

Since I'm just working with IPv4 right now, I can check the subnet in my HostName subclass, but it'd be nice to get this as a builtin--I'm not familiar with the rules for IPv6.

[seancfoley]

"hosts that cannot be routed to" is a vague concept. We would need something concrete and well-defined. This library has a number of categorizations of addresses according to specific criteria (like "isPrivate", "isZero", "isLocal", and others), but does not make judgements about those addresses, like whether a given address is routable in a specific subnet.

Also, HostNameParameters and IPAddressStringParameters are intended for parsing, to indicate what address formats are permitted and so on. They are not intended to make judgements about addresses after parsing. For that, just use methods on the IPAddress and HostName objects themselves.

[bherw]

Thanks for the quick reply!

Yeah, it is a bit vague, and really if one wants to be sure, it's context specific: some addresses are going to be connectable/routable from one subnet but not from another. That kind of thing seems outside the scope of this library, or at least of HostName validation.

What I'm thinking of is some criteria that we know is always true, regardless of context. E.g., the 0.0.0.0/8 IPs are not valid destination addresses per RFC1122 3.2.1.3, so we can always infer that such an address is not a valid destination without needing to do a lookup.

Right now, I'm doing something pretty basic, just adding a subclass to make a HostName that follows some additional validation rules so whenever I pass this class around I have something I know follows the rules I want, but I feel like maybe I'm using the library in an unintended way since I immediately ran into a StackOverflowException when overriding validate. It's pretty easy to work around, but makes me think there might be a simpler way of using HostName.

public class DestinationHostName extends HostName {
	static public final HostNameParameters PARAMETERS;
	static private final IPAddress ZEROS = new IPv4Address(0, 8);

	static {
		val parametersBuilder = new HostNameParameters.Builder()
				.allowEmpty(false)
				.allowPort(false)
				.allowService(false)
				.setEmptyAsLoopback(false);
		parametersBuilder.getAddressOptionsBuilder()
				.allowAll(false)
				.allowEmpty(false)
				.allowMask(false)
				.allowPrefix(false)
				.setEmptyAsLoopback(false);
		PARAMETERS = parametersBuilder.toParams();
	}

	static public boolean isValid(HostName hostname) throws HostNameException {
		try {
			new DestinationHostName(hostname.toString());
			return true;
		} catch (HostNameException e) {
			return false;
		}
	}

	private boolean startedValidation;

	public DestinationHostName(String hostname) throws HostNameException {
		super(hostname, PARAMETERS);
		validate();
	}

	@Override
	public synchronized void validate() throws HostNameException {
		// getAddress calls validate, so we need to avoid the circular call loop
		if (startedValidation) {
			return;
		}
		startedValidation = true;
		super.validate();

		if (this.isAddress()) {
			val address = this.getAddress();
			if (ZEROS.contains(address)) {
				throw new HostNameException(this.toString(), "IP range is not routable");
			}
		}
	}
}