Something similar is going to be needed in the data network, particularly as more traffic shifts to Internet Protocol. Typical IP routing is chaotic, allowing subsequent packets to take different routes across the network. Moreover, it's not built for reliability; IP's "best effort" philosophy takes a shoulder-shrugging approach to long delays and lost packets.

In addition, the optical network can be opaque to the rest of the network, a problem telephones never had to face. For example, all-optical cross-connects won't have any way of knowing what kind of traffic sits inside the light waves they're switching. Standards for telling the crossbar what to do are being formulated now, one example being the Optical Link Interface, an early signaling proposal being shopped to standards bodies such as the Internet Engineering Task Force (IETF) and the International Telecommunications Union.

The communication between the crossbar and the network is considered essential. "If we're selling photonic switches and there isn't something like this, the whole thing could go down," said Rajiv Ramaswami, vice president of system architecture at Nortel Networks.

Likewise, generalized multiprotocol label switching (GMPLS) and the Optical Internetworking Forum's user network interface (UNI) are being considered as ways for outlying routers to negotiate paths through the optical-networking core. Vendors seem to agree that both protocols are necessary, as G-MPLS reveals more information about network topology than carriers want to divulge.

New signaling standards aren't without a price, however. At the recent Opticon trade show, Equipe Communications Corp. engineer David Boland warned against the extra expenses that GMPLS might incur, because the standard will require extra microprocessor cycles and new software. Boland and others believe that plain MPLS will handle many of the problems that G-MPLS is intended to solve.

In crude terms, a packet-particularly if the node has to do nothing other than forward the packet to the next node. By contrast, basic IP forwarding requires each node to examine a packet's destination address and determine which direction to send the packet from there. MPLS has drawn massive attention among equipment vendors and service providers, leading to dozens of potential standards being discussed in the IETF. Ideas include using the header for security, prioritization or even to set up connections through the network.

Simple MPLS applications include traffic engineering, using the label to add priority levels to IP packets. Another early idea for MPLS has been to create virtual private networks-although it's up for debate whether IP routing should be involved, as explained in Riverstone Networks' article in this section. MPLS likewise has promise as a means of delivering Ethernet-based VPNs and other services, as described in an article by Appian Networks.

And because MPLS provides a way to send information from one node to another, it could become the great moderator between IP, ATM and frame relay, as described by Polaris and Alcatel engineers in this section. Questions still face MPLS as it moves into real-world networks: Polaris' article discusses how the protocol might be used alongside signaling standards such as GMPLS.

The articles in this section deal primarily with MPLS, but it's the combination of MPLS and new signaling standards that will help carriers derive revenues from the optical network.