Summary

  • RFC 1346 framed a shared-link bargain in opposite directions: an institution might demand a guaranteed minimum, while its network manager might demand a maximum on that institution's use.
  • Phil Jones also identified a seller's incentive to withhold customer control tools when the alternative was buying a larger link. Price could influence use; it could not allocate packets by itself.

Suppose the file transfers of one institution fill a shared international link just as another institution needs its terminals to respond. The second institution can buy a wider access line. That purchase may do nothing to restrain the first institution at the shared bottleneck. Phil Jones put this predicament in RFC 1346 in June 1992, when the question of sharing intercontinental network capacity was still being argued in terms of tools, budgets and institutional power.

Jones began with an unwelcome observation: traffic tended to fill the capacity available, and a line-speed upgrade carried no guarantee of durable relief. An IT service manager could tell the network manager that Department X deserved more capacity than Department Y, or that an interactive terminal session should outrank a file transfer. The equipment and contract still had to turn that preference into a rule at the contested link. A faster pipe at one edge was a poor substitute for a decision at the point of contention.

The two parties wanted different guarantees from the same link. An institution sharing capacity with larger partners might ask for a minimum share, so its users could not be swamped. The network manager might impose a maximum on each partner, perhaps in relation to what that partner paid. Both demands sound numerical. Jones pointed out that the number's unit was still unsettled. Was the share a width of access link, a rate of traffic, priority at busy hours, or something else the parties could measure and enforce? Until they chose a parameter, a promised slice was a negotiation phrase rather than an operating rule.

The hierarchy made the politics sharper. If the network manager could cap the institution, the institution would reasonably ask to divide its own allocation among departments and users. Otherwise authority stopped at the provider's boundary. The institution carried the cost of internal contention but lacked the controls needed to settle it. A network-wide rule without delegated local tools was governance by invoice.

The missing controls also blocked cooperation between network managers. Jones described a manager who might let another use a link occasionally as backup. Both could benefit. The owner still feared that backup traffic would swamp ordinary use, and lacked a way to bound it. A shared spare path was worth less when accepting a neighbor's traffic meant surrendering control of the service already promised to one's own users.

Then came the commercial sting. Jones wrote that a provider might have little interest in giving customers controls if their absence encouraged or obliged a customer to purchase more capacity. He also saw the reverse possibility: a provider with useful controls could win business. The memo described an incentive, not a proven account of any named carrier. But it located a real design conflict. The party able to sell an upgrade could be the same party deciding whether the buyer received tools to manage the old link.

Billing did not close that gap. A charge could influence behavior; unless use was prepaid and cut off at a threshold, it could not enforce one department's priority at the moment packets competed. Jones listed possible handles—source or destination address, application protocol, priority marking, X.25 throughput class—and admitted the practical tool set was thin. Static line splitting could reserve capacity, but it could also sacrifice the value of sharing it. Capacity unused at one hour vanished; the next institution's busy hour might begin across the ocean.

RFC 1346 ended by asking for studies, not by naming a winning protocol. Its most durable question was commercial as well as technical: who gets to define a share, who can enforce it at the bottleneck, and who can subdivide it after the bill has been paid? A bigger link answered only one of those questions.

Source: RFC 1346, Resource Allocation, Control, and Accounting for the Use of Network Resources, sections 1–3; RFC Editor record.