First National Bank of Chicago, as trustee of a real estate investment fund, brought this suit in federal district court against the Comptroller of the Currency to challenge his refusal to allow the bank to distribute individual properties owned by the fund to withdrawing investors in lieu of either cashing them out or giving them proportional interests in the fund’s entire portfolio of properties. Do not confuse this fund with a “Real Estate Investment Trust” (REIT), also a method of pooling real estate investments but one with special conditions not claimed to be satisfied here and an emphasis on tax avoidance that is irrelevant here because the participants in the fund are tax-exempt employee benefit plans. William A. Kelley, Jr., “How to Qualify a Real Estate Investment Trust (Part 1),” 6 Practical Tax Lawyer 31 (1991); Gregory W. Goff, “REIT Revival— An Overview of Real Estate Investment Trusts as an Investment Vehicle,” 39 Major Tax Planning, ch. 19 (U.S.C. Law Center Tax Institute 1987). This is a bank-managed “collective investment fund,” also known as a “common trust fund,” which has been helpfully defined as “a group of investments managed by a bank in accordance with a written plan on behalf of several individual fiduciary accounts whose assets are lawfully contributed to the fund.” William P. Wade, “Bank-Sponsored Collective Investment Funds: An Analysis of Applicable Federal Banking and Securities Laws,” 35 Business Lawyer 361 (1980). See also Investment Company Institute v. Camp, 401 U.S. 617, 621-22, 91 S.Ct. 1091, *136228 L.Ed.2d 367 (1971); Investment Company Institute v. Conover, 790 F.2d 925, 928 (D.C.Cir. 1986). Our fund happens to invest in real estate rather than in securities but that is an idiosyncrasy of this fund rather than a part of the definition of a collective investment fund. National banks are empowered to manage collective investment funds and engage in other fiduciary activities by virtue of 12 U.S.C. § 92a(a).

The Comptroller based his refusal to permit the bank to compensate withdrawing investors in the proposed manner on two regulations of his office governing the fiduciary powers of national banks. One provides that distributions to a withdrawing investor in a collective investment fund “may be made in cash or ratably in kind, or partly in cash and partly in kind.” 12 C.F.R. § 9.18(b)(6). The Comptroller interprets this to bar the form of distribution proposed by the fund, under which a withdrawing participant could be made to accept an individual parcel (or parcels) of real estate in exchange for his shares in the fund rather than receiving a proportional interest in all the fund’s properties. The other regulation provides that a fund that invests in real estate or other assets that are not readily marketable may require one year’s prior notice of withdrawals from it— but no more. 12 C.F.R. § 9.18(b)(4).

Institutional Real Estate Fund F was established in 1972 and has 45 participants — all substantial, sophisticated employee benefit plans, whose investments in the fund range from $50,000 to $100 million. In 1990 the fund had $59 million in cash and cash equivalents plus some sixty commercial real estate properties having an aggregate appraised value of $490 million. Yet the fund was worth less than it had been. The commercial real estate market had faltered as the decade of the 1980s came to a close. The value of the fund’s real estate holdings had declined by 10 percent, and withdrawal requests by investors in the fund had increased to the point where, by December 1989, the requests exceeded the fund’s cash. The bank suspended further withdrawals and distributions. It could have sold some of the fund’s properties to raise additional cash to fund withdrawals, but it didn’t want to do this, because of what it considered to be the depressed state of the market for commercial real estate. In March 1990, with fifteen withdrawal requests, aggregating $135 million, pending, the bank proposed to the participants a restructuring of the fund whereby any investor that wanted to make an immediate withdrawal would receive “in-kind distributions of the whole ownership in certain real properties.” Of the 45 participants, 37, owning in the aggregate 80 percent of the shares in the fund, agreed to the proposal.

The bank asked the Comptroller to rule that the proposal did not violate either of the regulations we have quoted, or in the alternative to grant a waiver from the regulations. The Comptroller wrote the bank that in his view the proposal violated both regulations and that he would not waive either of them. This lawsuit ensued. The district judge decided it on the pleadings— and rightly so, for although nominally an injunctive proceeding, in reality this is a proceeding to review agency action, and such proceedings are (with immaterial exceptions) decided on the record compiled before the agency. Cronin v. United States Department of Agriculture, 919 F.2d 439, 443-44 (7th Cir. 1990). The judge held that the regulation governing the form of distributions does not bar the proposed restructuring of the fund, but that the one-year limitation on notice of withdrawals does. Both parties have appealed.

There is jurisdictional underbrush to be cleared away before we can address the merits. First is the question of our appellate jurisdiction. The district court did not enter a separate judgment order, as required by Rule 58 of the Federal Rules of Civil Procedure. For when the judge ruled, a third count in the bank’s complaint was unresolved, so the judgment was not final and a Rule 58 order would have been premature. Later the parties agreed to drop the third count and an order dismissing it was duly entered on the docket sheet. But the clerk who entered it evidently did not realize that the order dismissing the re*1363maining count made the earlier judgment final.

Although the parties should have realized this and asked the judge to enter a Rule 58 order, the absence of the order is not fatal to our jurisdiction. The applicable statute requires a final judgment, 28 U.S.C. § 1291 (“final decision” is the statutory term, but the cases usually speak of the “final judgment” rule), and the requirement is jurisdictional, and therefore not waivable. But the statute does not say that the final judgment must be in the form prescribed by Rule 58. Compliance with the rule makes it easier to determine whether there is a final judgment, but if we know from other sources that there is one, we have jurisdiction. Bankers Trust Co. v. Mallis, 435 U.S. 381, 98 S.Ct. 1117, 55 L.Ed.2d 357 (1978) (per curiam); F. & H.R. Farman-Farmaian Consulting Engineers Firm v. Harza Engineering Co., 882 F.2d 281, 283 (7th Cir. 1989); Soo Line R.R. v. Escanaba & Lake Superior R.R., 840 F.2d 546, 549 (7th Cir. 1988).

Which is not to say that Rule 58 is one of those useless technicalities that bedevil and confuse the law. Compliance with it in this case might have avoided considerable confusion—might have told us for example who actually won in the district court (and therefore should be the appellant) and who lost (and therefore should be the appellee). Cf. American Interinsurance Exchange v. Occidental Fire & Casualty Co., 835 F.2d 157 (7th Cir. 1987). Remarkably, this is quite unclear. One interpretation of the judgment is that the losing party in the district court was the bank, because the court refused to vacate the Comptroller’s refusal to permit the proposed restructuring to go forward. On this interpretation, only the bank should have appealed. The Comptroller should have defended the judgment as appellee, Byron v. Clay, 867 F.2d 1049, 1050-51 (7th Cir. 1989), albeit on a different ground from the district court’s. Different because the Comptroller believes that his interpretation of the form-of-distribution regulation is correct and bars the restructuring, but that the one-year notice limitation, on the basis of which the district court upheld his refusal to permit the restructuring to proceed, has no independent significance. The reason is that if the form of distribution proposed by the bank is valid, the bank is agreeable to making the distributions authorized by it within one year of a request for withdrawal.

It could well be argued that given the Comptroller’s position on the significance of the one-year regulation, his victory in the district court was Pyrrhic and might better be interpreted as a loss. For by rejecting his interpretation of the regulation governing the form of distribution, the district court made the one-year regulation in effect moot as applied to this case. The regulation bites only if the bank is required to cash out the withdrawing participants, for then it would have to sell off properties in a hurry, and it doesn’t want to do this. If a withdrawing participant can be made to accept a whole-property distribution in kind, the bank is happy to make that distribution within one year of a request for it.

A party is allowed to appeal from a judgment nominally or for that matter substantially in his favor, provided only that it is adverse to him in some respect. La-Buhn v. Bulkmatic Transport Co., 865 F.2d 119, 121-22 (7th Cir. 1988); Disher v. Information Resources, Inc., 873 F.2d 136, 138-39 (7th Cir. 1989). The judgment of the district court was adverse to the Comptroller in a most important respect, for it upheld his action on a ground that would require him to rescind that action upon request by the bank. If he had not appealed, the bank would have resubmitted the request for restructuring, the Comptroller (not persuaded by the district judge’s opinion) would have turned it down on the ground that the form-of-distribution regulation barred it, the bank would have sued once more in the district court, and the court would have reentered its order. There would be an endless merry-go-round. This is a practical reason for allowing the Comptroller to appeal.

Although it does not matter in this case whether only the bank, or only the Comptroller, or both, could appeal, it could matter in another case. Litigants in cross-*1364appeal situations should be mindful that unless they are careful they may forfeit their right to all appellate review by standing on what they suppose to be their right to file a cross-appeal. If a timely notice of appeal is filed by one party, any other party may file its own notice of appeal within 14 days after that. Fed.R.App.Pro. 4(a)(3); Seafarers Int’l Union v. NLRB, 895 F.2d 385, 386 (7th Cir. 1990). But this is provided the first party had a right to appeal. If he did not, the second party must file within the normal time limit. Hameetman v. City of Chicago, 776 F.2d 636, 640 (7th Cir. 1985). If that is 60 days (as it was here, for there is a federal-government party, Fed.R.App.P. 4(a)(1)), and the first (the invalid) notice of appeal was filed on the fifty-ninth day and the second two days later, there would be no appellate jurisdiction. That is not a problem here, however, because both parties filed their notices of appeal within the time for filing a direct appeal as distinct from a cross-appeal.

So the appeal(s) is (are) before us, but there is a question, surprisingly little discussed in the case law, whether the Comptroller’s letter ruling is final agency action and therefore challengeable in a judicial review proceeding, or a mere advisory opinion, offering the bank the Comptroller’s interpretation of his regulations but leaving the bank free to go ahead with its restructuring of the fund and to face the consequences, if any, of its defiance. A letter doesn’t look much like final, formal agency action, and New York Stock Exchange, Inc. v. Bloom, 562 F.2d 736, 740-43 (D.C.Cir. 1977), held that a letter ruling by the Comptroller was not, while stressing that the form of the ruling should not matter. Id. at 740. See also First National Bank v. Smith, 508 F.2d 1371, 1373 n. 8 (8th Cir. 1974). In the present case, however, the bank requested not advice, perhaps on a purely hypothetical course of action, but permission to go forward with a concrete proposal that it had already put to the fund participants for their approval— permission based either on the Comptroller’s acceding to the bank’s interpretation of the regulations or on his waiving them. The Comptroller turned the bank down. He did not — the point that was stressed in Bloom, and that distinguishes that case from this one — offer a merely tentative view. Nor did he address hypothetical possibilities, or request additional information before making a definitive ruling. He said flat out that the bank could not restructure the fund as it wanted to do.

If this was the equivalent of the denial of a license or other permit, it was appealable. FTC v. Standard Oil Co., 449 U.S. 232, 239 n. 8, 101 S.Ct. 488, 493 n. 8, 66 L.Ed.2d 416 (1980); City of Anaheim v. FERC, 692 F.2d 773, 781-82 (D.C.Cir. 1982). The objection to the equation is the bank’s own argument that the instrument which established Institutional Real Estate Fund F and was approved by the Comptroller when the fund was set up already authorizes the proposed form of distribution, so that no further approval is required. If so, the bank can proceed with the distribution, albeit at some legal risk. It doesn’t need permission.

But this hardly seems a realistic view of the bank’s situation. The instrument allows distributions to withdrawing participants in the fund to be in cash or in kind, or partly in one and partly in the other, but it does not mention whole-property distributions. The instrument also does not use the word “ratably,” but it does provide that it “shall be administered in conformity with rules, regulations and rulings, if any, issued by any duly authorized governmental agencies,” and the Comptroller’s form-of-distribution regulation is one of those regulations and it was in force in 1972 when the fund was established. In these cloudy circumstances, for the bank to make so controversial a distribution without an authoritative determination of its right to do so would expose it to enormous potential liabilities for breach of trust, which is why trustees traditionally are allowed to seek direction from the court before embarking on a controversial course of action; only in this case it is the Comptroller rather than a court that supervises the trustee’s activities. Realistically, what the bank sought from the Comptroller and was refused was not advice, but permission, and if the bank *1365cannot challenge the denial of that permission it will have no practical opportunity to obtain a judicial interpretation of the regulation because it will have to fold its tent in the face of the Comptroller’s disapproval.

This construal of the agency’s action is especially persuasive with respect to the Comptroller’s refusal to waive the regulations. That surely was not advice; it was, indisputably, final agency action; and we can think of no reason it could not bring with it into the reviewing court, by analogy to pendent appellate jurisdiction, Cement Division v. City of Milwaukee, 915 F.2d 1154, 1158-59 (7th Cir. 1990); Asset Allocation & Management Co. v. Western Employers Ins. Co., 892 F.2d 566, 569 (7th Cir. 1989); Patterson v. Portch, 853 F.2d 1399, 1403 (7th Cir. 1988), closely connected nonfinal rulings such as the Comptroller’s interpretations of the regulations, if they were nonfinal. In any event we could not evaluate his refusal to waive the regulations without construing them.

Having satisfied ourselves, at last, with respect to jurisdiction in both the district court and this court, we can turn to the merits. The bank’s principal argument is a technical but not a negligible one. The predecessor regulation to section 9.18(b)(6) provided that “when participations are withdrawn from a Common Trust Fund distributions may be made in cash or rat-ably in kind, or partly in cash and partly ratably in kind.” 12 C.F.R. § 206.17(6), 2 Fed.Reg. 2976, 2978 (1937) (“Regulation F”) (emphasis added). The word we have italicized was dropped in 1963, when, as an incident to Congress’s transferring regulatory authority over national banks’ trust activities from the Federal Reserve Board to the Comptroller of the Currency, Regulation F was replaced by the present regulation, the only change being the deletion of the second “ratably.” The Comptroller concedes that the deletion was deliberate— it wasn’t just a slip-up in the process of repromulgation. The bank argues that so long as the fund adds some cash to its whole-property distributions, those distributions are authorized by the regulation because the requirement of ratable distribution applies only when a distribution is entirely in kind.

The argument disregards the deference that courts owe to agencies’ interpretations of their own regulations — a point made by the Supreme Court with reference to investment regulations of the Comptroller of the Currency in Investment Company Institute v. Camp, supra, 401 U.S. at 626-27, 91 S.Ct. at 1097. See also Federal Reserve System v. Investment Company Institute, 450 U.S. 46, 56-57 and n. 21, 101 S.Ct. 973, 981-82 and n. 21, 67 L.Ed.2d 36 (1981); Lyng v. Payne, 476 U.S. 926, 939, 106 S.Ct. 2333, 2341, 90 L.Ed.2d 921 (1986); United States v. Baxter Healthcare Corp., 901 F.2d 1401, 1407 (7th Cir. 1990); Homemakers North Shore, Inc. v. Bowen, 832 F.2d 408, 411-12 (7th Cir. 1987). It also creates an irrational distinction in the regulation. Suppose that among the fund’s properties are two worth $9.9 million and $10 million respectively, and a participant in the fund wants to make a $10 million withdrawal. Under the bank’s interpretation, if it gives the withdrawing participant the $10 million property it violates the regulation, but if it gives the participant the $9.9 million property plus $100,000 in cash it does not. What sense could that make?

The bank argues that if forbidden to make whole-property distributions, either it will have to liquidate — in what is widely believed to be a depressed market — a large part of its portfolio, in order to raise cash to fund withdrawal requests; or it will have to give withdrawing participants fractional shares in all the properties in the portfolio, resulting in divided ownership and, in turn, high costs of transactions. Suppose fifteen participants withdraw all or part of their investment. Under “ratable” (proportional) distribution, each of the sixty properties would have sixteen owners — the fifteen withdrawing participants, each of whom would receive shares in each of the fund’s properties, plus the fund itself. The more owners a piece of property has, the more costly it is to sell, improve, or otherwise transact with regard to it. This insight underlies the right of a tenant in common to a partition of the property, the Rule in Shelley’s Case, the presumption *1366(about which we wrote recently in Penn Central Corp. v. U.S. Railroad Vest Corp., 955 F.2d 1158, 1159-60 (7th Cir. 1992)) that the grant of a railroad right of way is an easement rather than a fee simple and therefore ceases, restoring undivided ownership, when the railroad discontinues service over the right of way, and many other doctrines of property law.

There is a second argument against forbidding the bank to make whole-property distributions. The value of a participant’s interest in the fund upon withdrawal depends upon the appraised value of the properties owned by the fund. In a down market, such as we now have in commercial real estate in this country, market value may be below appraised value (unless there is frequent reappraisal). If so, cashing out a withdrawing participant on the basis of the appraised value of his share of the fund might confer a windfall — which might in turn set off a race to withdraw. Whole-property distribution avoids this problem by giving the withdrawing participant property valued at the appraised value but presumably having a lower market value, rather than cash equal to the appraised value.

The second argument is easily met by pointing out that the bank is free to have its portfolio of properties reappraised in light of the current market situation. The first argument disregards the downside risk of whole-property distribution. Almost the entire idea behind a real estate investment fund is to allow investors to obtain a diversified investment in real estate, something a single investor unless truly huge would find very difficult to do because of the lumpiness of real estate compared for example to corporate securities. Investment Company Institute v. Conover, supra, 790 F.2d at 928; Wisconsin Real Estate Investment Trust v. Weinstein, 781 F.2d 589, 598 (7th Cir. 1986); cf. Central National Bank v. U.S. Dept. of Treasury, 912 F.2d 897, 902 (7th Cir. 1990); GIW Industries, Inc. v. Trevor, Stewart, Burton & Jacobsen, Inc., 895 F.2d 729 (11th Cir. 1990). Diversification is lost as soon as the investor’s interest is transformed from an undivided interest in a large and presumably diversified portfolio of real estate holdings into an interest in a single property. Ratable distribution preserves diversification, albeit with higher transaction costs, while cash distribution preserves diversification because the recipient of cash can forthwith invest it in some other diversified investment vehicle. In contrast, the recipient of a single building or other single chunk of real property must sell it before he can realize cash with which to make another diversified investment, and in the meantime he has an undiversi-fied investment. And every time a property is withdrawn from the portfolio and given to a withdrawing participant, the diversification of the portfolio declines, to the prejudice of the remaining participants.

The force of these objections should not be exaggerated. The investors in Institutional Real Estate Fund F are large and sophisticated and for all we know may be able to offset any loss of diversification as a result of their (or other participants’) receiving a whole-property distribution by making adjustments elsewhere in their investment portfolios. (But of course as sophisticated investors they presumably relied on the terms of the original trust instrument, which was silent on whole-property distribution.) Moreover, they may not want the fund to be forced into either liquidating a large part of its portfolio or encumbering title to the properties in the portfolio with the fractional interests of the withdrawing participants. Withdrawing participants themselves might, all things considered, prefer to receive whole properties. The lopsided vote in favor of the restructuring proposal so suggests. And although, as we have said, the bank is incorrect in arguing that the trust instrument itself clearly authorizes the bank to make whole-property distributions, the instrument’s very vagueness is some evidence that the participants did not wish to limit the bank’s distribution choices as narrowly as the regulation as interpreted by the Comptroller does — though a contrary possibility is that the instrument was drafted in light of the Comptroller’s consistent interpretation of the regulation.

*1367The issues of finance and economics raised by the bank’s forcefully presented arguments could be debated at greater length, but the important point for us is that they are arguments against the regulation rather than against the Comptroller’s interpretation. For the regulation explicitly and unequivocally requires ratable distributions in kind when the distribution is wholly in kind rather than partly in kind and partly in cash, yet the objections to forbidding whole-property distributions are independent of whether there is some cash or no cash. What is true is that a whole-property distribution entirely in kind is unlikely, because it would require that the fund just happen to have a property whose appraised value was exactly equal to the amount that the participant wanted to withdraw. But the bank makes nothing of this point, and perhaps this is right because the participant could always scale down (or up) his withdrawal request to the appraised value of some properties or properties owned by the fund.

So the reality is that the bank considers the regulation irrational but rather than challenging its validity asks us to create a loophole that would enable it to evade the regulation by adding a bit of cash to every distribution in kind. It says the cash component of the distribution would have to be more than de minimis. What could that mean? More than a dollar? More than a thousand dollars? More than a million dollars? As a matter of fact the restructuring proposal does not require the distributions to be partly in cash and partly in kind. If the value of the distributed property is less than the value of the withdrawing investor’s interest, the bank must make up the difference in cash. But if the value of the distributed property exceeds the value of the withdrawing investor’s interest, the investor must pay the fund the difference, in which case the only distribution by the fund is in kind. The bank itself does not take its loophole seriously.

Which is not to say that it is unthinkable that the second “ratably” was dropped from the regulation in 1963 in order to create this loophole. Who knows what backroom deals lie behind the change in the regulation? There is no indication of a deal, however, and it is at least as likely that the word was dropped because the draftsman incautiously thought it redundant. He may have thought that the first “ratably” would be understood as establishing that all distributions in kind must be ratable, and the following clause, dealing with distributions partly in cash and partly in kind, as merely waiving any requirement that a distribution be wholly one or wholly the other — but insofar as it has an in-kind component the distribution must be ratable. This reading is consistent with the semantics of the regulation, and though not compelled by them seems more plausible than the bank’s reading, given the absence of any indications of deliberate loopholing.

We said that the bank has not contested the validity of the regulation, but we add that if its suit is construed, as perhaps it should be, as an oblique attack on that validity, it must fail too, because the criterion is not whether we agree with the regulation but whether it is reasonable. (Besides the cases cited earlier, see Morales v. Yeutter, 952 F.2d 954 (7th Cir. 1991), and Wisconsin Electric Power Co. v. Reilly, 893 F.2d 901, 907 (7th Cir. 1990).) Our earlier discussion should make clear that it is.

The bank has other arguments, but only one has sufficient merit to require discussion. The Comptroller’s regulations provide that “funds ... held by a national bank as fiduciary may be invested collectively, to the extent not prohibited by local law.” 12 C.F.R. § 9.18(c)(5). The bank, interpreting “local law” to include ERISA (the Employee Retirement Income Security Act, 29 U.S.C. §§ 1001 et seq.), understands this regulation to mean that the Comptroller must allow the bank to manage the fund consistently with the “prudent man” standard of ERISA (all the fund participants are ERISA plans). And the bank regards that standard as requiring it, in the circumstances that it encountered at the end of 1989, to make whole-property distributions. This is not of course what the regulation says, and it’s a pretty tortured way of arguing that the Comptroller should *1368not interpret the form-of-distribution regulation to make an investment fiduciary act imprudently. We agree with that argument, but we do not think that the Comptroller’s interpretation does compel, or for that matter authorize, imprudent conduct. Suppose the real estate market really is so “depressed” today that it would be imprudent to sell real estate in order to raise cash with which to pay off withdrawing investors in a real estate investment trust. Free-marketers may challenge the assumption, arguing that the market price, and not some real estate appraiser’s appraisal (even if fully current), is the best, perhaps the only, measure of value. That is indeed a powerful argument with respect to a liquid market, such as the market in stocks traded on major stock exchanges — a market in which there are plenty of willing buyers and sellers, so that it is always possible to make a transaction that will reflect the value estimations of a number of market participants. There are some illiquid markets, though, Central National Bank v. U.S. Dept. of Treasury, supra, 912 F.2d at 902, and in them transaction prices will not always reflect market values, at least if “market” is defined to require some minimum number of potential transactors. If the author of this opinion decided to value his watch by soliciting bids from his two law clerks, he might get a less reliable valuation than from an appraiser. Likewise, if Institutional Real Estate Fund F decided to value a large office building by auctioning it off on an hour’s notice, the price it received would not be a “market” price in any meaningful sense of the word. While normally “the real estate appraisal process depends on sales of comparable properties and other market data,” “there have been few recent property sales in some markets, and bid and ask price may differ considerably.” Wayne E. Etter, “Appraising in Difficult Markets,” 6 Real Estate Center J. 14 (Jan. 1992). What is the value of a property when the owner is asking $3 million and a potential buyer is offering $2 million and there are no recent sales of comparable properties?

Illiquidity is what is behind the Comptroller’s one-year regulation. A real estate investment fund is not required to sell properties the instant it receives a withdrawal request that exceeds its cash reserves; it can take up to a year to find a buyer. Given this leeway it is difficult to see how the bank in our case could be thought imprudent if it decided to sell over the course of a year a large part or for that matter the entirety of its real estate portfolio. To fault it for selling in a depressed market would be to commit the fallacy of supposing that a prudent investor is one who can time market turns — a fallacy that investment law has at last rejected, prodded by modern finance theory. John H. Langbein & Bruce A. Wolk, Pension and Employee Benefit Law 539-40 (1990), and references cited there; Hamilton v. Nielsen, 678 F.2d 709, 712-13 (7th Cir. 1982). But if it should be imprudent to sell now, perhaps because, as the bank argues, though without substantiation, the savings and loans crisis has resulted in the placing on the market all at once of so much commercial real estate as to make it nearly certain that the bank could realize higher prices if it waited more than a year, all that this would mean is that instead of selling properties the bank should distribute fractional interests in them to withdrawing investors. That is not an ideal solution, because it makes such properties harder to sell. But whole-property distribution is not ideal either. Given a choice between comparable evils, neither could be pronounced imprudent by a court. The bank cites no case that condemns a fiduciary as imprudent for failing to make a whole-property distribution, and it would violate the supremacy clause for a state court to punish a fiduciary for complying with a valid regulation of the Comptroller of the Currency. There can be no violation of ERISA itself because that statute is explicit that it shall not be construed to invalidate or impair any federal regulation. 29 U.S.C. § 1144(d). And from all that we have said it should be plain that the Comptroller did not abuse his discretion in refusing to grant a waiver of the regulations.

We conclude that the ruling of the Comptroller denying the bank’s request to be *1369allowed to restructure the fund is valid in its entirety. The judgment of the district court is modified accordingly, and as modified is affirmed.

Modified and Affirmed.