EASTERBROOK, Circuit Judge.

Manus, Inc., gave the Los Angeles office of Merrill Lynch, Pierce, Fenner & Smith, Inc., a check for $647,250 payable to Merrill Lynch’s order. The check was drawn on Devon Bank in Chicago. Merrill Lynch immediately deposited the check with Crocker National Bank in Los Angeles; a clearing house presented the check to Devon for payment at 9:30 a.m. on Wednesday, August 1, 1979. The clearing house and Devon provisionally settled for the check immediately. Under § 4-301(1) of the Uniform Commercial Code, Devon had to decide no later than midnight of the next banking day whether to make final payment. See Ill.Rev.Stat. ch. 26 ¶ 4-301(1). Devon gave notice of dishonor at 4:22 p.m. on August 3: If this is too late, Devon is liable on the check even though Manus cannot cover the instrument. The district court thought the dishonor timely, 654 F.Supp. 506 (N.D.Ill.1987), and granted summary judgment to Devon in this diversity litigation.

The initial question is whether Devon gave notice of dishonor before the deadline on midnight of the “banking day” after it received the instrument. Under § 4-104(l)(c) of the UCC, a “ ‘banking day’ means that part of any day on which a bank is open to the public for carrying on substantially all of its banking functions”. On Wednesday, August 1, 1979, Devon’s lobby was closed to the public. It offered services, essentially limited to deposits and withdrawals, at a walk-up window. No one could open an account or arrange for a loan; so far as the record reveals, no one could draw down a line of credit previously arranged. Merrill Lynch observes that on Wednesdays Devon processed checks and made inter-bank loans, but neither these nor related activities made it “open to the public” for “substantially all of its banking functions”.

Devon is an Illinois bank, and a “bank” in Illinois is “any person doing a banking business whether subject to the laws of this or any other jurisdiction.” Ill.Rev. Stat. ch. 17 11302 (1986). So a person doing a “banking business” but not subject to anyone’s laws is not a “bank”, but the statute does not illuminate on “banking business”. Perhaps the statute uses a circular definition because the elements of banking are not particularly obscure. Making loans is a necessary part of “banking”; consider the definition of a “bank” in the Bank Holding Company Act, the only *1007 federal statute defining the term: “an institution ... which both (i) accepts demand deposits or deposits that the depositor may withdraw by check or similar means for payment to third parties or others; and (ii) is engaged in the business of making commercial loans.” 12 U.S.C. § 1841(c)(1)(B) (1987); see Board of Governors v. Dimension Financial Corp., 474 U.S. 361, 106 S.Ct. 681, 88 L.Ed.2d 691 (1986). Banks are financial intermediaries, facilitating transactions between those who want to lend and those who want to borrow. Cf. Arthur Allen Leff, The Leff Dictionary of Law, 94 Yale L.J. 1855, 2127 (1985). Devon, which was open to the public for only the deposit side of the banking business on August 1, 1979, was not open for “substantially all” of the services of a bank. Its services on August 1 were less extensive than those offered by a “nonbank bank” for purposes of the Bank Holding Company Act. Devon’s walk-up window may have been a “branch bank”, for both state and federal law define branches as places where deposits are received or money lent. 12 U.S.C. § 36(f); Ill.Rev.Stat. ch. 17 ¶ 302 (1986). One of these is not “substantially all” of the bank’s functions, however. The district court properly resolved this question by summary judgment. It would unacceptably disrupt commercial relations to put to a jury, case-by-case, the question whether a given day was a “banking day”. Billions of dollars in transactions must be processed by every midnight deadline, and everyone has an interest in having this time defined with precision. The record supplies enough information to make decision possible. Devon’s midnight deadline was 11:59 p.m. on Friday, August 3, 1979.

The midnight deadline is only the outside limit, however. Section 4-301(1) allows a bank to return an item if it acts “before it has made final payment (subsection (1) of Section 4-213) and before its midnight deadline” (emphasis added). Section 4-213(1) says that a settlement becomes final “when the bank has done any of the following, whichever happens first”. The only subsection we need consider is § 4-213(l)(e), which provides that payment becomes final when the bank has “completed the process of posting the item to the indicated account of the drawer, maker, or other person to be charged therewith”. Section 4-109 defines the process of posting, to which we return after stating some undisputed facts.

Manus, the maker of the check, had a subsidiary, Cash Reserve Management, Inc. Cash Management maintained an account in Boston. Manus gave its check to Merrill Lynch on July 26; on July 27 Ma-nus deposited in Devon a check for an identical sum of which Cash Management was the maker. Devon promptly submitted that check for payment. Devon places a “hold” of three or four business days on uncollected funds. The Manus check was presented for payment on August 1, the fourth business day (the fifth if Devon counted Saturday, July 28).

When Devon receives a bundle of checks from its clearing house, its computers tally the checks to ensure that the clearing house has debited Devon the correct amount. During the evening, reader/sorter machines read the account code on each check and compute the balance in each active account; a computer compares the balance and activity information with information the bank maintains to facilitate the decision whether to pay checks. The computer prepares, by the morning of the next business day, several reports for the bank’s staff. One report lists checks that have caused overdrafts in the account; another report lists checks that are subject to stop payment orders; a third report lists accounts in which uncollected funds are essential to cover the latest checks; there are more. The morning of the second business day, Devon returns most of the checks that appear on these lists — though its staff may elect to pay some of them. The bookkeeping department stamps checks “paid” and photographs them. Devon then examines the signatures on substantial checks. If the signature appears genuine (or if the bank elects not to examine the signature), Devon places the check in the customer's *1008 file. This process usually is completed in the afternoon.

Manus’s check was processed in the ordinary course. The account contained about $1.2 million, more than enough to cover the check. About $650,000 of this represented the Cash Management check deposited on July 27. Devon’s computer treated these as “collected” funds because the check had been deposited four or more business days ago. The uncollected funds reports of August 1 and 2 do not flag the Manus check. Devon verified the signature and placed the Manus check in the file during the afternoon of August 2. There it remained until 4:10 p.m. on August 3, when Continental Illinois National Bank told Devon by telephone that Cash Management’s bank in Boston had dishonored the check of July 27. At 4:22 p.m. Devon gave telephonic notice of dishonor of the Manus check. Crocker Bank resubmitted the Manus check, which was dishonored a second time; Manus was placed in receivership on August 28.

Merrill Lynch, which prefers collecting from a solvent Devon Bank to standing in line as one of Manus’s creditors, maintains that Devon “completed the process of posting the item” within the meaning of § 4-213(l)(c) during the afternoon of August 2, when it placed the check in the file. Devon had carried out all the steps in its ordinary process and planned to do nothing further. The process was free from operational error; no steps had been omitted, no judgmental blunders made along the way. Devon replies that it does not intentionally pay checks written against uncollected funds, to which Merrill Lynch responds that Devon made a business judgment to pay checks written against instruments that had been on deposit for four business days. Devon applied that rule to Manus’s check, and the belated return of the item may show that four days was too short but does not undermine the conclusion that “the process of posting” had come to an end.

The district court sided with Devon, 654 F.Supp. at 509-10, relying on § 4-109, which provides:

The “process of posting” means the usual procedure followed by a payor bank in determining to pay an item and in recording the payment including one or more of the following or other steps as determined by the bank:

(a) verification of any signature;

(b) ascertaining that sufficient funds are available;

(c) affixing a “paid” or other stamp;

(d) entering a charge or entry to a customer's account;

(e) correcting or reversing an entry or erroneous action with respect to the item.

Devon completed its ordinary steps, including each of (a) through (d), but the court concluded that § 4-109(e) gives a bank the privilege to dishonor a check until the midnight deadline. To return the item is to “reverse” the entry. As the court put it, “Devon’s returning the check ... demonstrated that the posting process was not completed” (654 F.Supp. at 510).

This reading of § 4-109(e) rips § 4-213(1)(c) out of the Uniform Commercial Code. Section 4-301(1) sets the midnight deadline as the last instant at which a check may be returned; § 4-213(1) lists four events that terminate the return privilege sooner. If the return of an item establishes that the “process of posting” was not completed, then § 4-213(1)(c) is meaningless. It is not beyond belief that statutes contain meaningless provisions, but a court should treat statutory words as dross only when there is no alternative. The Uniform Commercial Code is an uncommonly well drafted statute, with links among its provisions. Section 4-213(1) is there for a reason — to expedite the final settlement on checks, so that banks such as Devon Bank may make funds available to customers faster. The “midnight deadline”, and “deferred posting” in general, is a concession to the flux of paper with which any bank must contend. See Official Comment 1 to § 4-301. Section 4-213(l)(c) provides that final payment should not take any longer than the bank actually requires to process each item. That function would be defeated if the bank could reverse any *1009 posting under § 4-109(e) until the midnight deadline.

Perhaps § 4-213(l)(c) causes more trouble than it is worth. It potentially calls for a case-by-case inquiry into the details of posting; a bank may defeat its function by dragging out its normal processes so that they consume the entire period allotted by § 4-301(1); the drawee’s bank cannot rely on § 4-213(1)(c) to credit a customer’s account, because it does not know how long the drawer’s bank takes to post any given item. Considerations of this sort led the UCC’s Permanent Editorial Board, now at work on a Uniform New Payments Code, to propose the repeal of § 4-213(l)(c). See Uniform New Payments Code P.E.B. Draft No. 3 at 267-68 (1983) (calling the criteria of § 4-109 “unworkable in locating the precise time of payment”). Draft No. 3 was not adopted, though for reasons unrelated to §§ 4-109 and 4-213(1)(c); the Board continues to contemplate the problem. See Fred H. Miller et al., Commercial Paper, Bank Deposits and Collections, and Commercial Electronic Fund Transfers, 42 Bus. Law. 1269, 1288 (1987). Until the Board makes a final revision in the model UCC and states delete § 4-213(l)(c) — if that should occur — our job is to enforce the statute.

That § 4-213(l)(c) has meaning is reinforced by Official Example 3 to § 4-109:

A payor bank receives in the mail on Monday an item drawn upon it. The item is sorted and otherwise processed on Monday and during Monday night is provisionally recorded on tape by an electronic computer as charged to the customer’s account. On Tuesday a clerk examines the signature of the item and makes other checks to determine finally whether the item should be paid. If the clerk determines the signature is valid and makes a decision to pay and all processing of this item is complete, e.g., at 12 noon on Tuesday, the “process of posting” is completed at that time. If, however, the clerk determines that the signature is not valid or that the item should not be paid for some other reason, the item is returned to the presenting bank and in the regular Tuesday night run of the computer the debit to the customer’s account for the item is reversed or an offsetting credit entry is made. In this case ... there has been no determination to pay the item, no completion of the process of posting and no payment of the item.

This puts the “payment” of the check at the completion of the bank’s ordinary process, whatever that process may be. The check that passes the bank’s internal controls and is posted to the account is “paid”. None of the official comments suggests that a check that has been accurately handled in accordance with the bank’s ordinary procedure nonetheless may be dishonored any time before the midnight deadline.

Doubtless we must give § 4-109(e) meaning, just as we must leave some function for § 4-213(l)(c). The Supreme Court of Wisconsin thought the language of § 4-109(e) so “plain” that it overrode § 4-213(1)(c). West Side Bank v. Marine National Exchange Bank, 37 Wis. 2d 661, 669-72, 155 N.W.2d 587 (1968). That court read the language with this emphasis: “correcting or reversing an entry or erroneous action with respect to the item.” The “reversing an entry” language, the court thought, allowed the bank to dishon- or a check at any time before the midnight deadline, whether or not the processing had been completed without error. This reading is inconsistent with the official comments to § 4-109 and any plausible reason for making payment final on posting. As a result, the courts that have addressed the problem since 1968 have rejected West Side. Nelson v. Platte Valley State Bank & Trust Co., 805 F.2d 332 (8th Cir.1986); North Carolina National Bank v. South Carolina National Bank, 449 F.Supp. 616, 620 (D.S.C.1976); H. Schultz & Sons, Inc. v. Bank of Suffolk County, 439 F.Supp. 1137 (E.D.N.Y.1977); R. Hoag v. Valley National Bank, 147 Ariz. 137, 708 P.2d 1328 (1985). Students of the subject likewise disagree with the reading of § 4-109(e) proposed by West Side, although they are not entirely in accord on the meaning the section should take. E.g., William D. Hawkland & Lary Lawrence, 5 Uniform *1010 Commercial Code Series 323, 339-40, 345-46 (1984); James J. White & Robert S. Summers, Uniform Commercial Code § 16-4 & n. 38 (1980); Walter Malcolm, Reflections on West Side Bank: A Draftsman’s View, 18 Catholic U.L.Rev. 23 (1968); Note, Bad Checks and the U.C.C. —When is a Check Finally Paid?, 9 B.C. Ind. & Com.L.Rev. 957 (1968). We think it likely that the Supreme Court of Illinois would follow Nelson and Schultz rather than West Side.

Section 4-109(e) does not simply say that a bank may reverse an entry; the full text of the section says that it may correct or reverse an entry or erroneous action. It is not possible to divorce the “reversing an entry” language from the words immediately before and after. If we group the words this way — “(correcting or reversing) an (entry or erroneous action)” — the statute makes sense. The bank may correct (alter) or reverse (set aside completely) an entry that should not have been made or an “erroneous action” that does not involve an “entry”. This reading leaves a role for both § 4-109(e) and § 4-213(l)(c), and it also makes sense of the official comment to § 4-109, which states that when the bank’s ordinary process is completed before the midnight deadline, the check has been paid.

Nelson, Schultz, and the commentators on the payments process have stressed that the decision to pay a check has both mechanical and judgmental components. The examination of the signature and the determination that an account has sufficient funds are mechanical; the decision whether to permit an overdraft in the account is judgmental. Section 4-109 allows a bank to follow its ordinary processes for dealing with both of these. There is nothing magical about putting the instrument in the customer’s file. Suppose, for example, the bank mechanically puts all checks in customers' files and then makes random spot checks to verify signatures; that an item with a forged signature had to be pulled from the file to be returned would not prevent its dishonor. Or suppose the bank verifies signatures and puts the checks (stamped “paid”) in customers’ files before examining the computer printouts for stop payment orders. Again it would not be important that the bank had to remove the stopped check from the customer’s file. The alternative — holding all checks in stasis until each of the bank’s steps had been completed — would delay “final payment” for the checks as a group even longer, contrary to the purpose of § 4-213(l)(c). West Side may have been a case of this sort; if it was, we do not question its result even though its language was unduly broad. But none of this assists Devon. Its process — as Devon defines its process — had been completed by the afternoon of August 2. All of the check-specific steps, mechanical and judgmental, had been finished to the Bank’s satisfaction. The system functioned as it was supposed to. True, Devon may wish that it had told its computer to assume that checks take five rather than four business days to clear, but regret over a managerial judgment in the design of the check processing system is not a reason to dishonor a check after it has been posted to the account and finally paid.

Devon relies on Brown v. South Shore National Bank, 1 Ill.App.3d 136, 273 N.E.2d 671 (1st Dist.1971), for the proposition that Illinois follows West Side. Brown allowed a bank to return a check that had been stamped “paid” and placed in the customer’s file. It did not cite West Side and did not discuss the official comment to § 4-109. The disposition in Brown is entirely congruent with the one we believe appropriate. The drawer did not have a signature on file with the bank, so its processes apparently had misfired; the bank returned the check when it discovered the error. But the last paragraph of Brown contains this line: “The act of the defendant in returning the check to plaintiff’s bank demonstrated that the posting process was not completed, regardless of the entry on the ‘statement of account’ and the stamps, placed on the check.” 1 Ill.App.3d at 139, 273 N.E.2d at 673. This statement was unnecessary to the decision and is unreasoned; this dictum does not lead us to believe that the Supreme Court of Illinois will adopt the rationale of West *1011 Side. In Go-Tane Service Stations, Inc. v. Sharp, 78 Ill.App.3d 785, 33 Ill.Dec. 916, 397 N.E.2d 249 (2d Dist.1979), another appellate court of Illinois, without citing Brown, implied that the completion of signature checks and placement of the item in the file would prevent dishonor of a check, even if the customer then placed a stop-payment order and the bank returned the check before the midnight deadline. We need not resolve any tension between Brown and Go-Tane to conclude that the dictum in Brown is not the law in Illinois today.

Ill

Devon makes one last argument. Merrill Lynch sued Crocker Bank, claiming that it dallied in informing Merrill Lynch of the dishonor of the Manus check. The complaint in that suit states that had Merrill Lynch received timely notice of the dishonor, it could have obtained good funds from Manus to cover the check before Ma-nus entered bankruptcy. Devon says that this is an “admission” that Devon’s actions did not injure Merrill Lynch. Nonsense. Merrill Lynch had to mitigate any damages it suffered, and its pleading in the California case says that had it known of the problem, it could have mitigated. But it did not know, and therefore could not take the necessary steps. Whatever rights Devon may have against Crocker Bank, this pleading is hardly a reason why Merrill Lynch should lose.

Reversed.