dissenting in part:

This case presents three related legal issues stemming from a single set of facts. Balzer supplied to a contractor, S & S, certain items that Balzer reasonably believed would be used during the course of four U.S. Government construction projects. Defendant Fidelity had bonded the projects under four separate Miller Act bonds.

The first issue concerns the determination of general Miller Act liability when items are supplied to more than one bonded project. The district court held that the projects could be lumped together, so that if Balzer reasonably believed that its supplies would be used on any one of the projects, Fidelity would be liable. The majority, quite correctly, concludes that this was wrong; liability is determined on a contract-by-contract basis.

The second issue involves the statute of limitations. Here again, the majority rejects the district court’s single-pie theory, holding that each bonded contract is a separate event for purposes of triggering the statute. I fully agree with this portion of the majority opinion as well.

Oddly, the majority adopts the converse theory in resolving the third question: whether supplies are material or capital equipment. Without explanation, the majority concatenates the four projects, holding that supplies are material so long as the supplier reasonably believes they will be consumed in the projects as a group. It makes no sense to me to adopt a global approach on this issue while applying a balkanization rule to the first two. The three issues are obviously related; if consistent theories were applied, proof on the first two issues could easily go to the third. The majority’s unfortunate inconsistency will complicate problems of discovery and proof, and increases the danger of confusion by judge and jury. Moreover, the majority’s rule will create perverse incentives for contractors and Miller Act sureties. Worse, the majority’s approach is unique in the federal courts, creating a conflict with the law in at least three other circuits.

I

The Miller Act allows recovery to one who supplies “labor or material” to federal construction projects. 40 U.S.C. § 270b(a). It does not protect the supplier of capital equipment. The Miller Act provides an alternative remedy to those who would ordinarily be protected under a state mechanic’s lien. See F.D. Rich Co. v. U.S. ex rel Industrial Lumber Co., 417 U.S. 116, 122 , 94 S.Ct. 2157, 2161 , 40 L.Ed.2d 703 (1974). Because a lien cannot attach to federal property, the Miller Act provides a substitute by requiring that all federal construe *552 tion projects be bonded. Id. Suppliers of capital equipment are not protected by such bonds, because they cannot ordinarily get a mechanic’s lien. See, for example, Cal Civil Code § 3110 (West, 1974); Gordon Hardware Co. v. San Francisco & San Rafael R.R., 86 Cal. 620, 622 , 25 P. 125 (1890).

Those who supply material to a project without receiving full payment need the protection of a mechanic’s lien against the project, because an ordinary security interest is useless against material consumed in completion of the project. Suppliers of material are therefore also eligible for Miller Act protection. Suppliers of capital equipment, on the other hand, can protect themselves by taking a security interest in equipment that retains value at the end of the project. Suppliers of capital equipment are thus ineligible for mechanic’s liens or Miller Act protection.

It is with this logic in mind that we must examine the distinction between material and capital equipment under the Miller Act. Judge Haynsworth explained that distinction nicely in U.S. ex rel Sunbelt Pipe Corp. v. United States Fidelity & Guaranty Co., 785 F.2d 468, 470 (4th Cir.1986):

Sunbelt Pipe followed at least two other circuits in adopting this single-contract rule. See US. ex rel J.P. Byrne & Co. v. Fire Association of Philadelphia, 260 F.2d 541, 544 (2d Cir.1958) (“materials” are things “the use plaintiff and the construction company reasonably expected ... to have been substantially used up in the work under the contract”) (emphasis added); Massachusetts Bonding & Ins. Co. v. U.S. ex rel Clarksdale Machinery Co., 88 F.2d 388, 389 (5th Cir.1937) (“the supplying by sale to the contractor of equipment which should last over several jobs is not covered by the bond”) (emphasis added).

The key to this rule is that where a bond covers a particular project, there must be a reasonable expectation that the contractor will substantially consume the supplies on that project in order for the supplies to constitute material and therefore be recoverable under that particular bond. Supplies that retain value at the end of a project are not covered by the Miller Act because the supplier may protect himself by taking a security interest. The rule is also consistent with a common sense notion of capital: If a contractor can use something for more than one job, it is part of the capital of his enterprise; if he uses it up in a single job, it is material.

II

The majority adopts the contrary rule:

Majority op at 550 (emphasis added). I have no quarrel with part two of this rule; part one is nothing but trouble.

A. As noted, the other circuits that have looked at the material/capital problem have decided to treat each bonded project separately. In these circuits, if a piece of equipment is likely to survive a particular project it is capital equipment as to that project. This is an entirely manageable rule, one that treats all suppliers and bonding companies equally and predictably.

The majority here looks to the bonded projects as a group. One problem with this rule is that it makes it hard for a surety to calculate the premium it will charge for a particular bond. If the focus is on what the contractor can reasonably be expected to consume in a single project, the surety will, at the time it calculates its premium, have a rational basis for determining whether supplies are material, and therefore covered by the bond, or capital equipment, and therefore exempt from Miller Act protection. If the focus is, instead, on *553 some indeterminate group of present and future projects, the surety will be hard-pressed to figure out exactly what it is insuring. 1

Another difficulty with the majority’s rule is that the number of bonded projects will almost certainly vary from contractor to contractor, and this will have a direct effect on what is considered capital equipment. Thus, if a contractor has but one bonded project, only those items likely to be consumed on that project will be deemed to be material; everything else will be capital. An identically situated contractor who devotes more of his time to government work and has, say, ten bonded projects, will find that more of his equipment is deemed to be material because many more items will wear out during ten projects than during one.

Why should these two contractors be treated differently for purposes of determining whether the items they use are material or capital under the Miller Act? 2 I can think of no reason and the majority advances none. Surely the identity or business strategy of a contractor ought to have no bearing at all on whether the supplier of a widget can recover under a Miller Act bond. It makes far more sense to treat widgets as either material or capital regardless of how many government projects the contractor happens to have.

The majority’s rule is counterproductive. The more government projects a particular contractor undertakes, the more items in his inventory will be deemed material. In the extreme, the contractor who spends all of his time on government work will have no capital equipment at all because every piece of equipment — every cement mixer, every steamroller, every petroleum distillation plant — will be consumed during the course of the bonded projects “as a group.” 3 Bonding companies are not going to be oblivious to this fact; they will adjust their premiums accordingly. As a result of the majority’s rule, a contractor who does little government construction will be charged a far lower premium per project than one who does a great deal of government work. The size of the Miller Act premiums will, in turn, have to be figured into the bidding price of the contract. Experienced government contractors will be put at a competitive disadvantage as against occasional or novice contractors.

*554 This is a serious matter. The government has an interest in maintaining an ongoing relationship with contractors who perform well and who have managed to wend their way through the government’s Byzantine contracting procedures. Government contracting is a beast of a different color, and those who battle it and triumph should be handed a medal not a ball and chain. The majority’s rule will have the perverse effect of driving experienced contractors away from government work because their Miller Act premiums will become prohibitive. This makes no sense to me at all.

B. But we need not look quite so far into the future to see the majority opinion’s ill effects; some will show up on remand. As the majority has structured its opinion, the parties will have to try the case under two inconsistent theories: a contract-by-contract approach for general Miller Act liability and the statute of limitations, and an aggregate approach for the material/capital distinction. I can’t think of a way to complicate the problems of discovery or proof more perversely, or to confuse the judge and jury more thoroughly, than by adopting inconsistent theories on these closely related issues.

When crafting rules of law, appellate courts must attend to the very real problems of applying those rules in the crucible of litigation. Our law is complicated enough: We are almost to the point where federal jurors should be subject to dismissal for cause unless they answer yes to the question “Are you willing to disregard common sense in reaching a verdict?” Much of this complexity is unavoidable, as Congress passes laws that are increasingly complex and counterintuitive. See Stephen J. Adler, Can Juries Do Justice to Complex Suits?, Wall St J at B1 (Dec 21, 1989). But a substantial part of the problem can be laid at the doorstep of appellate judges who sometimes write opinions with all the common sense and practical wisdom of a law review article. But unlike law review articles, most of which, mercifully, are ignored by everyone involved in the workaday practice of law, impractical appellate opinions are like the proverbial crocodile in the bathtub: You can’t ignore them and they immensely complicate the task at hand.

To be sure, lawyers and district judges learn to adjust to whatever convoluted rule the court of appeals imposes on them, but this adjustment is not cost-free. The more complicated and counterintuitive the rule, the more court time, the more client money, the more witness time will be consumed in resolving the dispute. 4 Litigation, which already eats up a hugely disproportionate share of our productive resources, sucks yet another quantum of energy from the vital business of producing goods and services. In my view, this is a poor bargain. We will come to grief as a nation if we continue the current trend of robbing Peter to pay Paul’s lawyer. See P. Huber, Liability: The Legal Revolution and Its Consequences 153-71 (Basic, 1988).

Appellate judges have a vital role to play in regulating the degree of complexity— and therefore the cost — of our legal system. While we are bound by whatever Congress decrees in statutes, there are many occasions where we are called upon to craft a rule of law, either in interpreting the Constitution, or in performing our supervisory functions over the conduct of district court proceedings, or in common-law adjudication, or in drafting procedural rules; there are others. When performing that function, we must not give free reign to whatever rule happens to tickle our fancy-

Instead, we must, wherever possible, select the rule of law that advances fairness while diminishing the risk and cost of litigation; usually, they amount to the same thing. Thus, all things being equal, courts should favor a rule that respects the sanctity of written instruments, as such a rule promotes predictability and diminishes the *555 incentive to litigate. See Trident Center v. Connecticut Gen’l Life Ins. Co., 847 F.2d 564, 569 (9th Cir.1988). Similarly, courts should avoid creating vague new causes of action that invite parties and their lawyers to stalk lawsuits as if they were leprechauns. See Oki America, Inc. v. Microtech Int'l, Inc., 872 F.2d 312, 316 (9th Cir.1989) (Kozinski concurring). And, no less important, appellate judges should be mindful of the realities of litigation and avoid establishing rules of law that unnecessarily complicate the litigation process. See United States v. Balough, 820 F.2d 1485, 1491 (9th Cir.1987) (Kozinski & Hall concurring).

The approach the majority adopts runs afoul of this last axiom. With no explanation, and for no apparent reason, the majority creates a rule of law that requires the parties to litigate parallel legal issues by means of inconsistent legal theories. The majority no doubt prefers the rule it adopts, but is that a sufficient reason to impose this added burden on the district court, the lawyers and the parties? Even if it were the better rule — which I seriously doubt, for the reasons explained above— are we not also required to make sure it’s worth the candle? Absent some statutory imperative, which does not exist here, I suggest there is a heavy presumption in favor of adopting the consistent and more easily litigable rule; a rule focusing entirely on a single contract, or its converse, focusing on all bonded contracts as a group, would do fine for this purpose. To adopt a hybrid rule, which will send the parties and the district court chasing every which way is not, in my view, consistent with principles of sound judicial administration. To reach out for such a rule in the teeth of contrary authority from every other circuit that has addressed the issue is wholly incomprehensible. The majority may believe that it has reached a Solomonic solution, but, like Solomon, they may have merely reached a result that satisfies the court’s own sense of equity without genuine regard for whether its ruling makes sense. 5

Conclusion

Like every other court that adopts a broad and complicated rule, the majority predicts only modest consequences because of the “unique circumstances” of this case. Majority op at 549 n. 5. But there is nothing unique about this case; it’s a run-of-the-Miller Act dispute. I doubt that the majority’s disclaimer will deter many lawyers from pushing the majority’s maverick rule to the limits of its logic.

The majority’s rule runs counter to the rationale behind the Miller Act, and creates a conflict with at least three other federal circuits. Moreover, the rule will create needless confusion down the line. There is no explanation for any of this. When a court acts so contrary to ordinary reason, it must offer more justification than “[w]e do not demand that level of precision under the circumstances of this case.” Majority op at 550. I respectfully dissent.

. The majority contends that it would be just as difficult for a surety to determine before the fact whether supplies will be consumed in a single project as it would be to determine whether they will be consumed in many projects. See majority op at 550 n. 8. Not so. While there is uncertainty in any prediction, an insurer has a reasonable basis for determining what supplies are likely to be consumed on a specific project. The insurer can have no clue as to which supplies will be consumed in an indeterminate number of future projects. The former is a business judgment, the latter a crap-shoot.

. Here, the group of projects all happened to be bonded by the same Miller Act surety. Conceivably, then, the group the majority refers to consists of all projects bonded by the same surety. Under this construction, whether something is material or capital would turn on the identity of the surety; the contractor who uses a single surety will have far more material for Miller Act purposes than one who uses several different sureties. If this is what the majority has in mind, it should say so, although this methodology is based on an equally irrational distinction: whether the several projects are bonded by a single surety or several.

.Uncomfortable, perhaps, with the implications of its rule, the majority offers a limitation; "Obviously the above rule does not extend to items that accountants customarily regard as capital expenditures such as buildings, land, heavy construction equipment, etc.” Majority op at 550 n. 8. But there is nothing at all obvious about this. With the exception of land, each of the items the majority lists is clearly covered by the rule it adopts. How then will future litigants be able to tell what is covered by the rule and what is "obviously” not? The distinction appears to turn on what "accountants customarily regard." This introduces an entirely new issue for lawyers to argue about, adding yet another layer of complication and expense to what should be a relatively simple inquiry. Apparently, every Miller Act trial must now feature dueling accountants.

The majority's distinction is also inconsistent with its ruling in this case. The items here at issue, it will be remembered, are replacement parts for heavy earth-moving and rock-crushing equipment, the type of equipment the majority seems to say is "obviously” capital. Why would the parts to something that is capital be deemed material? The only rational answer relates to consumption, not to the opinion of accountants.

. This case provides a good example. If all the issues in a Miller Act case center around a single contract, discovery and proof will be so limited. But if, in deciding whether supplies are capital one must look at a whole string of projects, the number of documents and witnesses that must be examined is concomitantly multiplied.

. Solomon’s own reputation as a man of justice is probably overrated. His resolution of the famed maternity dispute rests on the presumption that the baby’s biological mother would be concerned about the life of the baby while the other woman would readily consent to its slaughter. Does this really make sense? As the story goes, the false mother was so grieved by the loss of her own baby that she stole that of another. 1 Kings 3:16-28. Would a woman in that situation countenance with indifference the killing of the very infant she had stolen to assuage her grief? In lieu of engaging in careful fact finding, Solomon may simply have handed the baby over to the woman who was clever enough to see through his bluff.

All of which is to say that Solomonic solutions may satisfy the Solomon in each of us, but do not necessarily reach the correct result. If Solomon’s experience teaches anything, it is that courts must be extremely wary of adopting rules of law that satisfy the court’s sense of justice but fail to take into account the realities of the dispute before them.