We have consolidated three petitions to review portions of two orders that the Environmental Protection Agency issued in December 1983: Approval and Promulgation of Air Quality Implementation Plans; Coke Batteries, Indiana, 48 Fed. Reg. 54599 (Dec. 6, 1983), and Approval and Promulgation of Implementation Plans; Indiana, 48 Fed.Reg. 55852 (Dec. 16, 1983). Those orders partly approved and partly disapproved the State of Indiana’s then latest revised plan for attaining compliance with the Clean Air Act, as amended, 42 U.S.C. § 7401 et seq. — what is called a “state implementation plan.” One of the two petitions, filed by two steel companies, complains about the agency’s disapproval in its order of December 6 of parts of the state plan that regulate coke oven doors and the “pushing” and “quenching” phases of coke production. The petition of these companies, as well as the petition of a third steel company, also complains about the agency’s decision in its order of December 16 to discontinue consideration of 1979 APC [Air Pollution Control]-9, a regulation proposed by the State of Indiana in an earlier revision of its state implementation plan and relating to the opacity of emissions from coke oven batteries. The companies want us to order the agency to act on that proposal — not let it (in their words) “pocket veto” the proposal by inaction. The third petition, filed by a citizens group, complains that the orders do not go far enough, and asks us to order the agency to undertake an additional rulemaking proceeding.
To understand this extremely complex case (which we shall ruthlessly simplify for the sake of intelligibility), you must know something about the scheme of the Clean Air Act and about the special problems of regulating air pollution caused by making coke. The Act, as comprehensively over*647hauled in 1970 (until then it had been quite toothless), contemplates a state-federal partnership for the control of air pollution. See Bethlehem Steel Corp. v. Gorsuch, 742 F.2d 1028, 1034-36 (7th Cir.1984); Currie, Air Pollution: Federal Law and Analysis (1981). The federal government establishes the permitted limits on particular pollutants, and then the states submit plans for achieving compliance with those limits (the “state implementation plans”). These plans are long documents, frequently revised, which contain many specific regulations, and which, before they can become effective, the Environmental Protection Agency must approve. If the EPA disapproves a particular regulation, it may either return the matter to the state for the preparation of a substitute regulation or it may, though only after notice and an opportunity for a hearing, promulgate its own regulation. Once a regulation is in force failure to comply with it exposes the polluter to heavy penalties. Additional major amendments to the Act were made in 1977, see Bethlehem Steel Corp. v. EPA, 723 F.2d 1303, 1304, 1305 (7th Cir.1983), but they are relevant to only a part of this case, so we postpone discussion of them till later.
Because of pervasive uncertainty regarding both the consequences of air pollution and the best methods of abating it, and enormous costs of abatement, the administration of the Clean Air Act has been afflicted by the related plagues of controversy, litigation, and delay, and nowhere more so than in the steel industry clustered in northern Indiana, on the southern shore of Lake Michigan. The production of steel is an extremely dirty process and the costs of bringing the industry into full compliance with federal air quality standards therefore immense. See H.R.Rep. No. 121, 97th Cong., 1st Sess. 4-9 (1981) (tabs. 1-3). Price increases that would pass on most of the costs of compliance to the consumers of steel would be difficult to make stick in the face of intense foreign competition. The industry would have to adjust by contraction — with all the social dislocations that the contraction of a major employer implies — beyond even the contraction that intense foreign competition has made inevitable.
A particular concern regarding the northern Indiana steel industry has been pollution from coke ovens. Coke of course is an input into the production of steel. It is made by heating coal in huge, almost airless ovens, arrayed in batteries. Because of the dearth of oxygen, the transformation through heating of coal into coke is not a combustion (burning) process, and hence the pollution that is produced as a byproduct of the process is not a product of incomplete combustion. • More to the point, this pollution is not vented in the usual way that the products of incomplete combustion are vented — through chimneys or smokestacks or the tailpipes of automobiles. Vented in that fashion it would be easier to deal with; or at least to measure, by putting a valve on the stack — and measurement is an essential part of determining compliance and hence of pollution control. The ovens are heated by a combustion process and the smoke produced by it is carried off in stacks, but this pollution is regulated in conventional ways and is not a point of controversy in this case.
The heating of the coal in the coke ovens throws off a dense yellow-brown mixture of gases called coke oven gas, which contains particulates and which can leak around the oven doors (as well as through cracked oven walls and piping) and thence out of the plant and into the atmosphere. The amount of leaking gas is greatest at the beginning of the heating process, since after a while the tars in the coke oven gas condense on the door, sealing it and thereby greatly reducing the leakage of gas. After the heating is completed the doors are opened and the coke “pushed” into rail cars for transfer to the “quench tower,” where the coke is sprayed with water to cool it off. Pushing and quenching throw up clouds of dust and steam (respectively) mixed with particulate matter, which again leak out of the plant and into the atmosphere.
*648Because heating, pushing, and quenching are intermittent, punctuating the daily life of the plant with bursts of gas, dust, and steam that cannot be vented through stacks (apparently it is not feasible to seal the whole plant and vent all of its emissions through stacks), the noncombustion emissions of a coking operation cannot easily be regulated just by setting an overall limitation on the amount of emissions. To assure compliance with such a regulation, monitoring instruments would have to be set up all around and above the steel plant and checked at frequent intervals in order to measure the total emissions from the coking process, and even then it would be necessary to separate those emissions from other emissions from the plant and from neighboring sources of pollution. And if the emissions could be measured, still it would be impracticable to achieve the desired limitation by installing a scrubber, catalyst, filter, or other control of the type familiar in regulating other “stationary source” pollution, precisely because the emissions are not vented through a pipe or stack, which could be fitted with such a device. Hence the strategy of opacity limitation and the separate regulation of oven doors, pushing, and quenching (some of the regulations employing pollution control devices, others not) — the aspects of the EPA’s December 1983 orders that the steel companies’ petitions have drawn in question.
The early history of Indiana’s struggle to regulate the opacity of noncombustion coking emissions is told in our opinion in Bethlehem Steel Corp. v. Gorsuch, cited earlier, and can therefore be recounted here very briefly. Indiana’s original state implementation plan, submitted to the EPA in 1972, contained a regulation, 1972 APC-3, which placed a 40 percent limitation (technically, No. 2 on the Ringelmann Chart, a measure of color rather than opacity but used as a proxy for the latter) on the opacity of emissions. The opacity of a plume of smoke or a cloud of dust or steam is the degree to which the plume or cloud can be seen through by an observer, the “plume observer” as he is called, usually a state pollution control inspector. Though a measure of unsightliness, opacity is not itself a form of pollution regulated by federal law; it is a proxy for particulate emissions; the more opaque the dust or smoke, the more particulates it can be expected to contain, on average. Because opacity sometimes is easier to measure than particulates, it is commonly used as a method for indirectly monitoring and limiting the amount of particulates in emissions, and was so used in Indiana’s 1972 APC-3. With regard to noncombustion emissions, however, which are the focus of concern in the present case, the regulation made failure to meet the 40 percent standard only prima facie evidence that the coke battery was in violation of the limitation (embodied in a companion regulation, 1972 APC-5) on particulates. The prima facie case could be rebutted by presenting acceptable evidence that the particulates limitation itself — the ultimate aim of the implementation plan so far as noncombustion emissions from coke batteries were concerned — was not being exceeded.
In 1974 Indiana replaced 1972 APC-3 with 1974 APC-3, which transformed what had been merely a prima facie case of exceeding the particulates limitation into a rigid 40 percent limitation on opacity, except that it allowed the limitation to be exceeded for 15 minutes a day. This regulation was submitted to the EPA and approved, but minus the 15-minute blow-off period, and in Bethlehem Steel Corp. v. Gorsuch we held that the agency could not do that; could not, under the guise of partial approval, put into effect without notice or hearing a regulation more stringent than the state had proposed. Meanwhile, after the agency had approved the 1974 version of APC-3 in its truncated form, Indiana had repealed 1974 APC-3 and replaced it with 1979 APC-9, which proposed to remove all opacity limitations on non-combustion coke emissions. When the EPA indicated that it probably would disapprove this proposal, too, the state submitted a new one, 1980 APC-3, which was intended to carry forward the opacity limi*649tation in the 1972 and 1974 APC-3’s but replace the 15-minute blow-off period that the EPA had disapproved in connection with 1974 APC-3 with a provision that compliance could be determined by averaging opacity over six-minute intervals; compliance every instant was not required. The previous APC-3’s had not indicated whether averaging was permissible for demonstrating compliance, but probably it was not. Without any specification of the period over which averaging is permitted, a regulation that permitted averaging would have no definite meaning. Since the previous regulations had not even mentioned averaging, they clearly had not specified the maximum period over which averaging was permitted. Presumably, then, it was forbidden.
The EPA in its order of December 16, 1983, which is one of the two orders before us today, disapproved 1980 APC-3 and suspended consideration of 1979 APC-9, which had never been formally disapproved by the EPA or withdrawn by the state. As a result of this disapproval the EPA’s truncated version of 1974 APC-3 — a strict 40 percent limitation on opacity with no blow-off period and (probably) no averaging — remained in effect. Then came our invalidation of the truncated version. The effect, under the EPA’s continuity doctrine, was to put back in force 1972 APC-3 — the last approved regulation of the opacity of non-combustion emissions — with its merely prima facie limitation. Approval and Pro: mulgation of Implementation Plans; Indiana, 49 Fed.Reg. 45178, 45179 (Nov. 15, 1984). The state’s version of 1974 APC-3, the version with the 15-minute blow-off period, was not restored because the EPA had never approved that version.
The EPA has acquiesced in our decision, and has begun a rulemaking proceeding (which is still in progress) to decide whether to reverse its previous disapproval of 1980 APC-3 and approve that regulation after all, as it might provide more control over emissions than 1972 APC-3. See Approval and Promulgation of Implementation Plans; Indiana, 49 Fed.Reg. 45178 (Nov. 15, 1984). All this happened after the filing of the petitioners’ briefs in this court, and the EPA argues that in light of these developments the steel companies’ challenge to its “pocket veto” of 1979 APC-9 is moot.
We do not think the challenge is technically moot, in the sense that our decision could have no effect on the parties. 1979 APC-9 is even more favorable to the companies than 1972 APC-3, because the latter contains some limitation on opacity (though maybe a weak one, depending on the feasibility of rebutting a prima facie case of having exceeded the particulates limitation), and the former no limitation. Therefore, if the EPA were to act promptly on 1979 APC-9 and approve it, or, if it disapproved it, were to be reversed and ordered to approve it, the companies would be under an even more favorable regime than they are now, when, as a result of the EPA’s disapproval of 1980 APC-3 (a disapproval that may, however, as we have noted, soon be rescinded) and our invalidation of the truncated 1974 APC-3, they are back under the regime of 1972 APC-3. Hence the relief the companies are asking for could make them better off, and that prevents their petitions for review on this point from being moot. There is also no doubt that the EPA’s refusal to act on the state’s proposal of 1979 APC-9 is a final order reviewable in this court. Indiana & Michigan Elec. Co. v. EPA, 733 F.2d 489, 490-91 (7th Cir.1984).
But to grant the remedy that the steel companies request of us, and hence to order the EPA to determine the validity of 1979 APC-9, would, even .if wrong there was, be a poor exercise of our equitable remedial discretion — and administrative review is a branch of equity, see, e.g., Mosey Mfg. Co. v. NLRB, 701 F.2d 610, 613 (7th Cir.1983) (en banc); Mobay Chemical Corp. v. Gorsuch, 682 F.2d 419, 427 (3d Cir.1982); Florida Power & Light Co. v. Costle, 650 F.2d 579, 590 (5th Cir.1981). The EPA has limited resources, in general and with respect to the problems created by coke batteries in northern Indiana, and *650we cannot believe that it would be a productive use of those resources to rake over the dying embers of 1979 APC-9. It is wholly unlikely that the EPA would approve a regulation that lifted all opacity limitations on noncombustion emissions from coking operations. It indicated its probable disapproval of 1979 APC-9 in the notice of rulemaking that it issued when the regulation was submitted to it for approval, and given the importance of opacity limitations in the control of air pollution caused by the noncombustion phases of producing coke (of which more shortly), we can be reasonably confident (without meaning to prejudge the issue) that the EPA will not change its mind on 1979 APC-9 or be forced to do so by a reviewing court, which is to say, by this court.
Moreover, although 1979 APC-9 is even more attractive to the companies than 1980 APC-3, they seem quite prepared to live with the latter, with its six-minute averaging; and the EPA has indicated in its notice of reopened rulemaking that it probably will approve 1980 APC-3, now that its gerrymandered 1974 APC-3 has been tossed out the window. In the end, of course, it may decide not to approve it, and then either the ball will be back in Indiana’s court (as has been the pattern in the regulation of Indiana’s coke batteries) or the EPA will try to draft a substitute regulation. Whatever it does, 1979 APC-9 will be left trailing in the dust; and it smacks almost of harassment of this harassed agency for the companies to continue pressing us — after the EPA’s acquiescence in our decision disapproving its version of 1974 APC-3, and after the EPA dismissed with prejudice its enforcement proceedings against the companies for refusing to comply with that regulation — to make the agency go back and reopen the proceeding on a six-year-old proposal and give that proceeding priority over everything else it is doing, on pain of punishment for contempt of this court. All this is not to condone the exercise by the EPA of a “pocket veto,” of regulation by delay; but on that our views are adequately stated in Indiana & Michigan Elec. Co. v. EPA, supra, 733 F.2d at 492, and need not be repeated here.
Still another reason for not getting involved with 1979 APC-9 is that 1980 APC-3 exempts coke batteries that are subject to 1981 APC-9, the other regulation under review in this case. Because the agency disapproved some provisions of 1981 APC-9, the exempting provision of 1980 APC-3 will not fully vest even if 1980 APC-3 is approved. Eventually, however, coke oven batteries will be governed entirely by the coke oven battery regulation, 1981 APC-9 or its successor, and then neither 1979 APC-9 nor 1980 APC-3 will have any but academic interest, at least so far as the coking operations of the steel companies are concerned.
And so we come to the EPA’s order of December 6, 1983, the other order before us, which the steel companies challenge insofar as it disapproved provisions of 1981 APC-9 governing oven doors, pushing, and quenching, as applied to “nonattainment areas” in Indiana. In its 1977 amendments to the Clean Air Act, Congress established new, more stringent requirements for areas of the country that had not yet attained compliance with the federal air quality standards. See Bethlehem Steel Corp. v. EPA, supra, 723 F.2d at 1305, 1308. Among these requirements is that the EPA make polluters in nonattainment areas comply with the national standards by December 31,1982, and that it “require, in the interim, ... such reduction in emissions from existing sources in the area as may be obtained through the adoption, at a minimum, of reasonably available control technology,” 42 U.S.C. § 7502(b)(3) — “RACT,” the inevitable if unlovely acronym for this requirement. The features of Indiana’s state implementation plan regarding oven doors, pushing, and quenching that the EPA disapproved it disapproved because they were inconsistent with RACT, and the steel companies’ first argument is that since 1982 has come and gone without achievement of the applicable air quality standards the requirement of complying with RACT has expired.
*651There is verbal support for this argument in the words “in the interim,” but if ever a literal reading disserved legislative intent, it is in this case. Indiana’s failure to achieve timely compliance with the national air quality standards makes it more rather than less needful that its polluters adopt reasonably available control technology. Congress’s idea, obviously, was not that pollution control would become weaker after 1982 but that until 1982 polluters would be required only to make progress toward attainment. They were to achieve it by the end of 1982 and thereafter remain in compliance using whatever measures had gotten them into compliance in 1982, or better measures if available. Having failed to achieve compliance on time they can scarcely be allowed to relax their efforts to achieve at least a belated compliance.
It is true that Bethlehem Steel Corp. v. EPA, supra, 723 F.2d at 1306-10, held that the EPA could not ignore the deadlines established by the 1977 amendments for classifying areas as nonattainment areas. The companies argue that, analogously, the EPA ought not be allowed to require compliance with RACT after the deadline for attainment has passed. But the analogy is a weak one. We held that the EPA could not impose the more stringent regulations authorized for nonattainment areas without following the statutory procedures for determining nonattainment, which included a deadline. We did not hold that when the target date for attainment comes and goes without attainment, the controls applicable to nonattainment areas expire.
So let us turn to the specific regulations that the EPA disapproved, beginning with the regulation of emissions from oven doors, which as we said leak coke oven gas, particularly at the beginning of the heating cycle, before the doors become sealed by the condensation of the tars in the gas. Indiana’s implementation plan provided that 10 percent of all the doors in each battery, plus 4 doors in the battery, would be allowed to leak; the EPA disapproved this, and said the most it would allow to leak in each battery would be 10 percent of operating doors, plus 4. So if there were 100 doors altogether (and therefore 50 ovens, because each has 2 doors) but the company was using only 50, Indiana’s proposal would allow leaks from 14 of the 50 doors (10 percent of all doors plus 4) but the EPA’s suggested revision would allow leaks from only 9 doors (10 percent of the 50 operating doors, plus 4).
It should be apparent from the way we have stated the issue that there are preliminary issues of mootness, ripeness, and finality, which we must of course resolve before considering the companies’ challenge to the disapproval on the merits. In disapproving Indiana’s proposal in the respects of concern to the steel companies the EPA sent the matter of regulating oven doors back to the drawing board in Indiana; and it is conceivable that Indiana will come up with a substitute proposal that will be agreeable to both the EPA and the companies. We do not think that this possibility makes the challenge moot or unripe, however. The ground of the EPA’s disapproval shows that if Indiana-wants to regulate oven doors it will have to do so on the basis of operating doors rather than all doors; and with oven doors a major source of noncombustion emissions from coking, it is hardly likely that Indiana will give up trying to regulate them, or will regulate them even less stringently than it proposed (for example by allowing 90 percent of operating doors to leak, rather than 10 percent of all doors plus 4, as it tried before); or that if the state does either of these things the agency will allow the state to get away with it. The EPA’s order has as a practical matter ruled out a mode of regulation quite favorable to the companies; and while in principle they can challenge this disapproval several years from now when some new regulation on oven doors is in place, by then the momentum of regulation may have carried the EPA and us beyond the point where it would be feasible to go back to the “cold doors” approach.
*652A more difficult question is whether the order is final. Doubt persists whether an order disapproving a proposed state implementation plan is final action within the meaning of the Clean Air Act, see Currie, supra, § 9.09, despite the 1977 amendments to section 307, as a result of which the courts of appeals have jurisdiction to review the EPA’s “action in approving or promulgating any implementation plan, ... or any other final action” by the EPA under subchapter I of the Act, 42 U.S.C. § 7607(b)(1), which is the subchapter in which the provision for disapproval of state implementation plans is found. But we need not resolve the issue here. The disapprovals of which the steel companies complain are contained in an order otherwise approving Indiana’s proposed revisions to its state implementation plan. The order considered as a whole is a final order and someone aggrieved by an aspect of it can challenge that aspect by asking us to set aside the order in whole or part. Public Service Co. of Indiana v. EPA, 682 F.2d 626, 637-38 (7th Cir.1982). It would not do to have piecemeal judicial review of a single order, with persons challenging the respects in which the order approved the state’s proposal being able to sue first and others having to await the completion of further proceedings before the agency. Particular issues might not be ripe because of the pendency of further proceedings on them, but ripeness is not a problem here, as we have said. Nor need we decide whether, if the order had approved Indiana’s proposed revisions in only one minor respect, and had disapproved the rest and sent the state back to the drawing board, the order might, considered as a whole, be so lacking in finality as to preclude judicial review (though within the literal scope of section 307 as amended in 1977), other than of the single respect in which the order had approved, and thus made indisputably final, the state’s regulation. That might be a case of the tail wagging the dog, cf. Abney v. United States, 431 U.S. 651, 663, 97 S.Ct. 2034, 2042, 52 L.Ed.2d 651 (1977); Kershner v. Mazurkiewicz, 670 F.2d 440, 448-49 (3d Cir.1982) (en banc), but this is not.
So we come to the merits of the oven-doors regulation. The companies point out with some logic that the amount of air pollution is no less if they do less coking, as distinct from doing the same amount but more carefully, i.e., with less leakage. It is becoming a familiar principle in the analysis of tort law that potential injurers can avoid accidents by either of two basic methods — by taking more care, or by reducing the amount of activity that they engage in that gives rise to accidents; in the traditional example of damage to crops from locomotive sparks, by having good spark-arresting equipment, or by running fewer trains per day. See, e.g., Shaved, Strict Liability versus Negligence, 9 J. Legal Stud. 1 (1980). The principle applies to pollution as well as to accidents. If a change in activity level is cheaper in the circumstances, the companies argue, they should not be penalized for preferring to make that change rather than to take greater care: operating fewer ovens, rather than sealing the doors of their operating ovens more tightly.
This is not a bad point, but its application to Indiana’s proposed rule (10 percent of all doors, plus 4 doors, may leak) is not obvious. The rule seems to reward companies that have the good fortune (in other settings the misfortune) to have a lot of unused capacity. They can pollute with impunity; a company with 100 doors (i.e., 50 ovens) that uses only 14 need do nothing to prevent those 14 (10 percent plus 4) from leaking. Moreover, the state’s rule conflicts with the principle of RACT (not necessarily the details: the companies argue that the EPA rejected the rule on the basis of some technical gibberish buried in 13,500 pages of RACT documentation on which the EPA relied without giving the state or the companies a chance to comment; this just is wrong — the rejection is based on the principle, not the particulars, of RACT). The principle is that polluters in nonattainment areas must use whatever control technology is reasonably available. The *653Indiana rule presupposes and the companies concede that even if a coke battery is running at full capacity, leakage can still feasibly be limited to 10 percent of the doors plus 4. This is done by beginning coking in the various ovens sequentially rather than simultaneously, so that only a few of them are at any one time in the initial stage of the heating cycle, where leakage is unavoidable. It follows that compliance with a 10 percent plus 4 rule confined to operating ovens is achievable too, since that rule equates to a 10 percent plus 4 rule for all ovens when all ovens are operating, as they may be. The companies argue that because leakage is unavoidable while the ovens are heating up, they can’t possibly comply on that basis. But if they had the good fortune to be operating at full capacity they would have to comply on that basis under Indiana’s rule — which they support.
Our conclusion is independent of the precise meaning of that elusive term, “reasonably attainable control technology.” The legislative history suggests that more than technical feasibility is involved; that economic feasibility is important too. “Whether a measure is reasonable also depends on the level of control to be obtained compared to the economic or social cost.” S.Rep. No. 127, 95th Cong., 1st Sess. 40 (1977). But the companies as we have said make no objection, economic or technical or otherwise, to a regulation (the “cold doors” regulation proposed by Indiana but rejected by the EPA) that, were they lucky enough to be running at full capacity, would have the identical effects as the alternative regulation preferred by the EPA; and that is a fatal objection to their challenge.
After the EPA issued the order under review in the present case — indeed, after argument in this court — it approved, for non-attainment areas in Kentucky, a 10 percent limitation on emissions from all coke oven doors, not just operating doors. See Approval and Promulgation of Implementation Plans; Kentucky; Removal of Conditions and Approval of Part D TSP Plan, 50 Fed.Reg. 41912, 43194-95 (Oct. 16, 1985). Missing, however, is the plus 4 feature of the Indiana proposal, so we are not prepared to conclude that the EPA’s action in the Kentucky proceeding demonstrates the irrationality of its rejection in this proceeding of a limitation based on all doors, especially since Indiana is a nonattainment area with regard to many categories of air pollution and Kentucky’s air pollution problems are less serious. See Designation of Areas for Air Quality Planning Purposes; Kentucky; Etc., 49 Fed.Reg. 40424 (Oct. 16, 1984).
This brings us to pushing. The state proposed that the companies be required to install a device designed to capture 90 percent of the particulate emissions. The EPA disapproved the proposal on the ground that design efficiency and operational efficiency may differ, and that there would be no way of determining what fraction of the particulate emissions was actually being captured by the device and therefore no way of determining whether the device was operating as it was designed to do. The agency suggested that instead the state come up with an opacity limitation for emissions from pushing.
Both halves of the EPA’s logic seem reasonable to us, and no more is required to put them beyond the power of judicial revision. Pushing creates great billows of dust, which are not vented. The control device would have to surround the coke ovens and rail cars like a tent to capture all of the emissions. In fact, devices called sheds or hoods exist for doing just this. But they are not panaceas; among other problems they increase the pollution to which the workers in the coking plant are exposed. If for these or other reasons a less encompassing control device is used, as is contemplated in this case, it will be impossible to determine whether the device is capturing 90 percent, or some much lower fraction, of the emissions. This is the basic reason for the use of opacity limitations and seems as applicable here as to the other stages of the coking process. Limiting opacity is a well recognized technique of pollution control and one especially ap*654ropos where as in the case of noncombustion emissions from coke batteries the determination of compliance with limitations stated in terms of a permitted amount of particulates is infeasible, or at least very difficult and expensive.
The problem of control is particularly acute with regard to quenching, because limiting the opacity of quenching emissions is of questionable feasibility (though attempted in the Kentucky regulation mentioned earlier, see 50 Fed.Reg. at 41915). Spraying cold water on hot coke throws up vast clouds which are entirely opaque, but, because consisting largely of steam, not so rich in particulates as the opacity of the clouds would suggest. Opacity is not in these circumstances a good proxy for particulate pollution. Indiana proposes instead to require a certain type of baffle above the quench tower and to limit the amount of dissolved solids in the water used to quench — the “make-up” water as it is called, which fills up with potentially polluting solids as it is used over and over again to quench the coke. The EPA approved the proposal except with regard to the provisions for sampling the make-up water and testing it for dissolved solids. Indiana’s proposal does not specify the frequency or location of the sampling or the method of testing the samples for dissolved solids.
These deficiencies are serious and warranted the EPA’s disapproving the proposed regulation in detail (as distinct from concept). How often the make-up water is sampled is critical to determining whether the companies are complying with the limit on dissolved solids, given the irregular nature of the coking cycle and the fact that the water is reused. Where it is sampled is also important. If it is sampled at too great a distance from where the water is actually sprayed on the coke, there is a danger of underestimating the amount of solids in the water when sprayed; for in traveling the remaining distance to the coke the water may pick up additional dissolved solids from other waste water; and some solids escape into the air when the water is vaporized. Finally, the EPA was certainly well within its rights in asking that Indiana agree to test for the amount of dissolved solids by a standard methodology published by the ASTM (American Society for Testing Materials) and not challenged, so far as we are aware, by the companies.
Having rejected the companies’ challenges we turn to the challenge mounted by the citizens group, which, as * it has members who live in Indiana and whose health may be affected by the air pollution from coking plants there, has standing. Sierra Club v. Morton, 405 U.S. 727, 92 S.Ct. 1361, 31 L.Ed.2d 636 (1972). The citizens group wants us to issue an order requiring the EPA, within six months, to issue new rules governing pollution from coking operations in nonattainment areas in Indiana, these rules to include: a fresh approach to the problem of coking emissions, having due regard for other types of air pollution; regulations of oven doors, pushing, and quenching to take the place of the disapproved parts of the Indiana implementation plan; and a regulation requiring the companies to monitor coking emissions by hiring plume observers to supplement Indiana’s corps of pollution enforcers. The citizens group wants the parts of Indiana’s implementation plan that the EPA approved to be allowed to go into effect but only as interim regulations which will be superseded when the EPA issues its new regulations at the end of the six months.
About the merits of these proposals we shall have nothing to say in this opinion, because we are convinced that we have no authority to issue the type of order that the citizens group seeks. The Clean Air Act disjoins judicial review of final action by the EPA, which as we have said section 307 places in the court of appeals, from judicial review of inaction by the agency, as to which section 304 provides that “any person may commence a civil action on his own behalf ... against the Administrator [of the EPA] where there is alleged a failure of the Administrator to perform any act or duty under this chapter *655which is not discretionary with the Administrator____ The district courts shall have jurisdiction ... to order the Administrator to perform such act or duty....” 42 U.S.C. § 7604(a). If the EPA’s refusal to undertake the rulemaking proceeding requested by the citizens group was the failure to perform a nondiscretionary duty, then exclusive jurisdiction to remedy that failure lies in the district court, while if the failure was a failure to perform a discretionary duty, it follows not that there is jurisdiction in this court but that there is jurisdiction in no court. When an agency has discretion as to whether or not to undertake rulemaking, the courts cannot tell it how to exercise that discretion. Agency action committed to the agency’s discretion is unreviewable in any court. 5 U.S.C. § 701(a)(2); see Heckler v. Chaney, — U.S. -, 105 S.Ct. 1649, 84 L.Ed.2d 714 (1985).
The argument of the citizens group is not that inaction in discretionary matters is final action for purposes of section 307 (review in this court), but that since as we have held the December 6, 1983, order is a final order by the EPA, it is reviewable in this court and only in this court and we have the power to grant any remedy that will cure a valid objection to the order. So th$ complaint is that the order does not go far enough and that we can order the agency to go farther.
It is true that if the citizens group were challenging the validity of the order because the EPA had failed to do something it was duty-bound to do, we would have jurisdiction. That was the Indiana & Michigan Elec, case, cited earlier. (Indeed, it is this case, so far as the steel companies’ challenge to the EPA’s tabling of consideration of 1979 APC-9 is concerned.) In that case two public utilities, in a forerunner of Bethlehem Steel Corp. v. Gorsuch, had challenged an order by the EPA approving one of Indiana’s air pollution regulations minus an averaging provision in the regulation that made it more lenient, as to which the EPA simply took no action; and we held that such a challenge was properly brought in this court. It was a challenge to final agency action taken on a full administrative record, so that the district court’s comparative advantage in compiling a record provided no argument for treating it as a section 304 case. See 733 F.2d at 490-91.
The citizens group is not challenging the validity of the EPA’s order. On the contrary, it wants the order to go into effect; it just wants it demoted as it were to an interim regulation that will give way when the EPA issues more comprehensive and ambitious regulations along the lines sought by the group. It is true that the EPA said in the order that it would not undertake the more ambitious program that the citizens group sought; and if that decision had tainted the remedial parts of the order in the group’s eyes then the group could challenge the order in this court. It would be just like the Indiana & Michigan case, or the virtually identical Kamp v. Hernandez, 752 F.2d 1444, 1454 (9th Cir.1985), or the part of this case where the steel companies challenge the agency’s discontinuance of proceedings on 1979 APC-9, as a result of which the companies will be under a more stringent regulation, 1972 APC-3. But so far as the citizens group is concerned, the EPA’s order of December 6 (or the one of December 16, for that matter) has no consequences it doesn’t like; at least it has no consequences that the group wants us to remove by refusing to enforce the order in whole or in part. The group is happy with the order for now; that is, as an interim order. It just wants more in the future, and to compel the agency to do more in the future it must proceed in the district court under section 304.
This is especially clear since we could not review the EPA’s determination not to go the route urged by the citizens group on the basis of the administrative record before us, which contains nothing on the feasibility of the EPA’s conducting the type of proceeding urged by the group, either within the next six months or within any other period. If this is a matter reviewable at *656all, it will require the compilation of a new administrative record that will disclose the reasons for and against a finding that a new rulemaking proceeding would be a feasible and fruitful undertaking, given everything else on the EPA’s crowded platter. See Currie, supra, § 9.11, at p. 9-33. But making a record is for the district court, not this court, to do. True, it would not be proper for even the district court to take evidence designed to lay a factual basis for (or against) the agency’s action. See Florida Power & Light Co. v. Lorion, — U.S. -, 105 S.Ct. 1598, 1607, 84 L.Ed.2d 643 (1985). But if, because of the highly informal nature of that action (a refusal to conduct a rulemaking proceeding), the agency’s reasons for the action are not disclosed, then a district court has the authority, and (as we do not) the practical ability, to compile a record limited to reconstructing, as distinct from supporting or refuting, the agency’s reasoning process. See, e.g., Edgewater Nursing Center, Inc. v. Miller, 678 F.2d 716, 718 (7th Cir.1982).
Considerations of remedy reinforce our conclusion that this matter has no business in this court, and this in two respects. First, by asking what remedy a complainant seeks, often one can more accurately characterize the nature of the right being asserted. The remedy sought by the citizens group is not to rescind or modify the rule that the EPA has adopted; it is to order the EPA to conduct a new, follow-on rulemaking proceeding. When all that is being complained of is a failure to undertake a new proceeding distinct from that which is under review, the complainant has stated a claim if at all only under section 304. Second, federal courts of appeals are not set up to exercise managerial or administrative functions effectively; for those functions a single judge, a district judge, not a panel of three or more court of appeals judges, is the efficient and appropriate tribunal; only in extremis will we allow ourselves to become a forum of original jurisdiction. The virtual demise of the three-judge district court, on which see Wright, The Law of Federal Courts § 50 (4th ed.1983), is practical proof of this point, if any is needed. We are not equipped to evaluate the monthly progress reports that the citizens group asks us to direct the EPA to submit or to pass on the inevitable request at the end of six months for an extension of time or the hardly less inevitable request that we institute contempt proceedings when and if the EPA defaults on our last extension. If these functions have been given to the judiciary to be performed at all they have been given to another level of the judiciary — to the district courts by section 304 of the Clean Air Act. As a matter of fact the citizens group has brought and is actively prosecuting such a suit. Citizens for a Better Environment v. Ruckelshaus, No. 80-C-0003 (N.D.Ill., E.D., filed Jan. 2, 1980).
The citizens group has submitted to us an unpublished order by the Sixth Circuit setting forth a timetable for compliance with a previous order of that court requiring both the EPA and Michigan’s counterpart agency to undertake various rulemaking proceedings. See Natural Resources Defense Council, Inc. v. EPA, No. 83-3027 (6th Cir. Oct. 1, 1984) (per curiam order). We are asked to infer from this that the Sixth Circuit is willing to undertake the same administrative duties that we would have to undertake if we granted the relief that the citizens group requests. We decline to draw the inference, because we know nothing of the circumstances that may have led the Sixth Circuit to do what it did in Natural Resources Defense Council. They are not disclosed in the two-page typewritten order that the citizens group has submitted to us and there are no published opinions in the case. It may be that we and the Sixth Circuit disagree on the scope of appellate competence to ride herd on an administrative proceeding but we do not know enough about the Natural Resources Defense Council case to know whether there is a real conflict. The circumstances of that case may be radically different from those of the present one.
Oljato Chapter of Navajo Tribe v. Train, 515 F.2d 654 (D.C.Cir.1975), held *657that section 307 provides the only route for judicial review of refusals by the EPA to revise performance standards, as well as judicial review of the promulgation of such standards; but that case is distinguishable. After the court of appeals in a section 307 action had upheld certain air-quality standards for new coal-fired electrical generators, Indian tribes that had not joined in the petition to review the promulgation of those standards asked the EPA to revise them, and when it refused brought a suit in district court under section 304. The court saw this as a fairly transparent effort to circumvent the prescribed route for judicial review; and not only were the courts of appeals intended to have exclusive jurisdiction but since the standard in question was a national one, one court of appeals, the court of appeals for the District of Columbia Circuit, had exclusive jurisdiction, so that the petitioners were trying to replace a single review proceeding in a specified court of appeals with district court proceedings all over the country. See 515 F.2d at 660-61. Here the citizens group, in bringing its section 304 action in the district court to compel the EPA to launch a new rulemaking proceeding, was not trying indirectly to undo the existing rule; the group likes the rule; it just wants the agency to promulgate an even stronger rule for future application. The Indian tribes in Oljato wanted a different rule; the citizens group here wants an additional rule; it is challenging not what the agency has done but what it has failed to do and a suit complaining of inaction can be brought only in the district court. In City of Sea-brook v. Costle, 659 F.2d 1371, 1373 (5th Cir.1971), also relied on by the citizens group, the court characterized the objectors as complaining “not only of what [the Administrator of the EPA] has done, but of what he has failed to do.” But in this case the citizens groups has no objection to what the EPA has done, in approving Indiana’s proposed revisions to its implementation plan in part and rejecting the revisions in part. . Its only objection is to the EPA’s refusal to launch a brand-new rulemaking proceeding.
The steel companies’ petitions for review are denied; the petition of the citizens group is dismissed.