IN RE GOOGLE REFERRER HEADER PRIVACY LITIGATION, Paloma Gaos; Anthony Italiano; Gabriel Priyev, individually and on behalf of all others similarly situated, Plaintiffs-Appellees, v. Melissa Ann Holyoak; Theodore H. Frank, Objectors-Appellants, v. Google, Inc., a Delaware corporation, Defendant-Appellee.
No. 15-15858
United States Court of Appeals, Ninth Circuit.
Argued and Submitted March 13, 2017. San Francisco, California. Filed August 22, 2017.
869 F.3d 737
For these reasons, I respectfully concur in the majority opinion, but I would also grant habeas relief based on Brecht‘s footnote nine.
Kassra P. Nassiri (argued) and John J. Manier, Nassiri & Jung LLP, San Francisco, California, for Plaintiffs-Appellees.
Donald M. Falk (argued) and Edward D. Johnson, Mayer Brown LLP, Palo Alto, California; Daniel E. Jones, Mayer Brown LLP, Washington, D.C.; Randall W. Edwards, O‘Melveny & Myers LLP, San Francisco, California; for Defendant-Appellee.
Before: J. CLIFFORD WALLACE, M. MARGARET McKEOWN, and JAY S. BYBEE, Circuit Judges.
Partial Concurrence and Partial Dissent by Judge WALLACE
OPINION
McKEOWN, Circuit Judge:
Google‘s free Internet search engine (“Google Search“) processes more than one billion user-generated search requests every day. This case arises from class action claims that Google violated users’ privacy by disclosing their Internet search terms to owners of third-party websites. We consider whether the district court abused its discretion in approving the $8.5 million cy pres-only settlement and conclude that it did not.
Background
In these consolidated class actions, three Google Search users—Paloma Gaos, Anthony Italiano, and Gabriel Priyev (collectively “plaintiffs“)—asserted claims for violation of the
The claimed privacy violations are the consequence of the browser architecture. Once users submit search terms to Google Search, it returns a list of relevant websites in a new webpage, the “search results page.” Users can then visit any website listed in the search results page by clicking on the provided link.
When a user visits a website via Google Search, that website is allegedly privy to the search terms the user originally submitted to Google Search. This occurs because, for each search results page, Google Search generates a unique “Uniform Resource Locator” (“URL“) that includes the user‘s search terms. In turn, every major desktop and mobile web browser (including Internet Explorer, Firefox, Chrome, and Safari) by default reports the URL of the last webpage that the user viewed before clicking on the link to the current page as part of “referrer header” information. See In re Zynga Privacy Litig., 750 F.3d 1098, 1102 (9th Cir. 2014) (explaining how “referrer headers” operate).1
The genesis of the plaintiffs’ complaints is the application of the search protocol, coupled with Google‘s “Web History” service, which tracks and stores account holders’ browsing activity on Google‘s servers. Following mediation, the parties reached a settlement, which they submitted to the district court for preliminary approval in July 2013. The settlement provided that Google would pay a total of $8.5 million and provide information on its website disclosing how users’ search terms are shared with third parties, in exchange for a release of the claims of the approximately 129 million people who used Google Search in the United States between October 25, 2006 and April 25, 2014 (the date the class was given notice of the settlement).
Of the $8.5 million settlement fund, approximately $3.2 million was set aside for attorneys’ fees, administration costs, and incentive payments to the named plaintiffs. The remaining $5.3 million or so was allocated to six cy pres recipients, each of which would receive anywhere from 15 to 21% of the money, provided that they agreed “to devote the funds to promote public awareness and education, and/or to support research, development, and initiatives, related to protecting privacy on the Internet.” The six recipients were AARP, Inc.; the Berkman Center for Internet and Society at Harvard University; Carnegie Mellon University; the Illinois Institute of Technology Chicago-Kent College of Law Center for Information, Society and Policy; the Stanford Center for Internet and Society; and the World Privacy Forum. Each of the recipients submitted a detailed proposal for how the funds would be used to promote Internet privacy.
After a hearing, the district court certified the class for settlement purposes and preliminarily approved the settlement. Notice was given to the class on April 25, 2014, via a website, toll-free telephone number, paid banner ads, and press articles. Thirteen class members opted out of the settlement, and five class members, including Melissa Ann Holyoak and Theodore H. Frank (collectively “Objectors“), filed objections.
Analysis
The settlement at issue involves a cy pres-only distribution of the $5.3 million or so that remains in the settlement fund after attorneys’ fees, administration costs, and incentive awards for the named plaintiffs are accounted for. Cy pres, which takes its name from the Norman French expression cy pres comme possible (or “as near as possible“), is an equitable doctrine that originated in trusts and estates law as a way to effectuate the testator‘s intent in making charitable gifts. Nachshin v. AOL, LLC, 663 F.3d 1034, 1038 (9th Cir. 2011). In the class action settlement context, the cy pres doctrine permits a court to distribute unclaimed or non-distributable portions of a class action settlement fund to the “next best” class of beneficiaries for the indirect benefit of the class. Id.
Here, the cy pres recipients were six organizations that have pledged to use the settlement funds to promote the protection of Internet privacy. We review for abuse of discretion the district court‘s approval of the proposed class action settlement. Id. In addition, because the settlement took place before formal class certification, settlement approval requires a “higher standard of fairness.” Lane v. Facebook, Inc., 696 F.3d 811, 819 (9th Cir. 2012) (quoting Hanlon v. Chrysler Corp., 150 F.3d 1011, 1026 (9th Cir. 1998)), cert. denied sub nom. Marek v. Lane, — U.S. —, 134 S.Ct. 8, 187 L.Ed.2d 392 (2013). Recognizing that, at this early stage of litigation, the district court cannot as effectively monitor for collusion and other abuses, we scrutinize the proceedings to discern whether the court sufficiently “account[ed] for the possibility that class representatives and their counsel have sacrificed the interests of absent class members for their own benefit.” Id.
I. Appropriateness of the Cy Pres-Only Settlement
As an initial matter, we quickly dispose of the argument that the district court erred by approving a cy pres-only settlement. Notably, Objectors do not contest the value of the settlement nor do they plead monetary injury. To be sure, cy pres-only settlements are considered the exception, not the rule. See Klier v. Elf Atochem N. Am., Inc., 658 F.3d 468, 474 (5th Cir. 2011) (explaining that direct distributions to class members are preferable because “[t]he settlement-fund proceeds, having been generated by the value of the class members’ claims,” are “the property of the class“); accord William B. Rubenstein, Newberg on Class Actions § 12:26 (5th ed. 2017). However, they are appropriate where the settlement fund is “non-distributable” because “the proof of individual claims would be burdensome or distribution of damages costly.” Lane, 696
The district court‘s finding that the settlement fund was non-distributable accords with our precedent. In Lane, we deemed direct monetary payments “infeasible” where each class member‘s individual recovery would have been ”de minimis” because the remaining settlement fund was approximately $6.5 million and there were over 3.6 million class members. Id. at 817-18, 820-21. The gap between the fund and a miniscule award is even more dramatic here. The remaining settlement fund was approximately $5.3 million, but there were an estimated 129 million class members, so each class member was entitled to a paltry 4 cents in recovery—a de minimis amount if ever there was one. The district court found that the cost of verifying and “sending out very small payments to millions of class members would exceed the total monetary benefit obtained by the class.”
To begin, the district court found that the amount of the fund was appropriate given the shakiness of the plaintiffs’ claims. Objectors do not contend that it would have been feasible to make a 4-cent distribution to every class member. Instead, they ask us to impose a mechanism that would permit a miniscule portion of the class to receive direct payments, eschewing a class settlement that benefits members through programs on privacy and data protection instituted by the cy pres recipients. Objectors suggest, for example, that “it is possible to compensate an oversized class with a small settlement fund by random lottery distribution,” or by offering “$5 to $10 per claimant” on the assumption that few class members will make claims. Our review of the district court‘s settlement approval is not predicated simply on whether there may be “possible” alternatives; rather, we benchmark whether the district court discharged its obligation to assure that the settlement is “fair, adequate, and free from collusion.” Lane, 696 F.3d at 819 (quoting Hanlon, 150 F.3d at 1027). If we took their objections at face value, Objectors would have us jettison the teachings of Lane. Objectors would also have us ignore our prior endorsement of cy pres awards that go to uses consistent with the nature of the underlying action. Nachshin, 663 F.3d at 1039-40.2
Likewise, we easily reject Objectors’ argument that if the settlement fund was non-distributable, then a class action cannot be the superior means of adjudicating this controversy under
II. The Cy Pres Recipients
We now turn to the crux of this appeal: whether approval of the settlement was an abuse of discretion due to claimed relationships between counsel or the parties and some of the cy pres recipients. We have long recognized that the cy pres doctrine, when “unbridled by a driving nexus between the plaintiff class and the cy pres beneficiaries[,] poses many nascent dangers to the fairness of the distribution process,” because the selection process may then “answer to the whims and self interests of the parties, their counsel, or the court.” Nachshin, 663 F.3d at 1038-39; see also Dennis v. Kellogg Co., 697 F.3d 858, 865 (9th Cir. 2012); Six (6) Mexican Workers v. Ariz. Citrus Growers, 904 F.2d 1301, 1308-09 (9th Cir. 1990). Due to these dangers, we require cy pres awards to meet a “nexus” requirement by being tethered to the objectives of the underlying statute and the interests of the silent class members. Nachshin, 663 F.3d at 1039.
Objectors suggest that the district court rubber-stamped the settlement, by “simply h[olding] that the Ninth Circuit and district courts have approved other all-cy pres settlements and class members effectively had no right to complain about the parties’ choice of compromise.” That characterization is unfair and untrue. And oddly, despite this claim, Objectors do not dispute that the nexus requirement is satisfied here.
The district court found that the six cy pres recipients are “established organizations,” that they were selected because they are “independent,” have a nationwide reach and “a record of promoting privacy protection on the Internet,” and “are capable of using the funds to educate the class about online privacy risks.” Although the district court expressed some disappointment that the recipients were the “usual suspects,” it recognized that “failure to diversify the list of distributees is not a basis to reject the settlement ... when the proposed recipients otherwise qualify under the applicable standard.” Accordingly, the district court appropriately found that the cy pres distribution addressed the objectives of the Stored Communications Act and furthered the interests of the class members. Previous cy pres distributions rest on this same understanding of the nexus requirement. See, e.g., Dennis, 697 F.3d at 866-67 (no nexus between false advertising claims relating to the nutritional value of Frosted Mini-Wheats® and charities providing food for the indigent); Lane, 696 F.3d at 817, 820-22 (nexus between Facebook privacy claims and charity giving grants promoting online privacy and security); Nachshin, 663 F.3d at 1039-41 (no nexus between breach of privacy, un-
Nonetheless, Objectors take issue with the choice of cy pres recipients, because Google has in the past donated to at least some of the cy pres recipients, three of the cy pres recipients previously received Google settlement funds, and three of the cy pres recipients are organizations housed at class counsel‘s alma maters. See In re Google Buzz Privacy Litig., No. C 10-00672 JW, 2011 WL 7460099, at *3 (N.D. Cal. Jun. 2, 2011). The Objectors point to a comment from the American Law Institute‘s (“ALI“) Principles of the Law of Aggregate Litigation which suggests that “[a] cy pres remedy should not be ordered if the court or any party has any significant prior affiliation with the intended recipient that would raise substantial questions about whether the selection of the recipient was made on the merits.” Principles of the Law of Aggregate Litig. § 3.07 cmt. b (Am. Law Inst. 2010) (emphasis added).4
The benchmark for “significant prior affiliation” is left undefined. Id. Of course it makes sense that the district court should examine any claimed relationship between the cy pres recipient and the parties or their counsel. But a prior relationship or connection between the two, without more, is not an absolute disqualifier. Rather, a number of factors, such as the nature of the relationship, the timing and recency of the relationship, the significance of dealings between the recipient and the party or counsel, the circumstances of the selection process, and the merits of the recipient play into the analysis. The district court explicitly or implicitly addressed this range of considerations.
We do not need to explore the contours of the “significant prior affiliation” comment because in the context of this settlement, the claimed relationships do not “raise substantial questions about whether the selection of the recipient was made on the merits.” See id. § 3.07 cmt. b.5 As a starting premise, Google‘s role as a party in reviewing the cy pres recipients does not cast doubt on the settlement. In Lane, we approved a cy pres-only settlement in which the distributor of the settlement fund was a newly-created entity run by a three-member board of directors, one of whom was defendant Facebook‘s Director of Public Policy. 696 F.3d at 817. We rejected the claim that this structure created an “unacceptable conflict of interest,” explaining that “[w]e do not require ... that settling parties select a cy pres recipient that the court or class members would find ideal” since “such an intrusion into the private parties’ negotiations would be improper and disruptive to the settlement process.” Id. at 820-21. Instead, we
Given the burgeoning importance of Internet privacy, it is no surprise that Google has chosen to support the programs and research of recognized academic institutes and nonprofit organizations. Google has donated to hundreds of third-party organizations whose work implicates technology and Internet policy issues, including university research centers, think tanks, advocacy groups, and trade organizations.6 These earlier donations do not undermine the selection process employed to vet the cy pres recipients in this litigation. The district court conducted a “careful [] review” of the recipient‘s “detailed proposals” and found a “substantial nexus” between the recipients and the interests of the class members. Notably, some of the recipient organizations have challenged Google‘s Internet privacy policies in the past.7 Most importantly, there was trans-
parency in this process, with the proposed recipients disclosing donations received from Google. Each recipient‘s cy pres proposal identified the scope of Google‘s previous contributions to that organization, and, unlike in Lane, explained how the cy pres funds were distinct from Google‘s general donations. See Dennis, 697 F.3d at 867-68 (casting doubt on the value of cy pres funds that a defendant “has already obligated itself to donate“). Citing Lane, the district court found that “[t]he chosen recipients and their respective proposals are sufficiently related so as to warrant approval; they do not have to be the recipients that objectors or the court consider ideal.”
The objection that three of the cy pres recipients had previously received cy pres funds from Google does not impugn the settlement without something more, such as fraud or collusion. See Rodriguez v. W. Publ‘g Corp., 563 F.3d 948, 965 (9th Cir. 2009). That “something more” is missing here. Indeed, the proposition that cy pres funds should not be awarded to previous recipients would be in some tension with our nexus requirements. As we have recognized, it is often beneficial for a cy pres recipient to have a “‘substantial record of service,‘” because such a record inspires
Finally, we reject the proposition that the link between the cy pres recipients and class counsel‘s alma maters raises a significant question about whether the recipients were selected on the merits. There may be occasions where the nature of the alumni connections between the parties and the recipients could cast doubt on the propriety of the selection process. But here, we have nothing more than a barebones allegation that class counsel graduated from schools that house the Internet research centers that will receive funds. The claim that counsel‘s receipt of a degree from one of these schools taints the settlement can‘t be entertained with a straight face. Each of these schools graduates thousands of students each year. Objectors have never disputed that class counsel have no ongoing or recent relationships with their alma maters and have no affiliations with the specific research centers. Nor did the district court simply accept this concession or put the burden on the Objectors. The district court appropriately considered the substance of the objections and explained why those challenges did not undermine the overall fairness of the settlement. See In re Pac. Enters. Sec. Litig., 47 F.3d 373, 377 (9th Cir. 1995). The court affirmatively analyzed the issue and was cognizant of the claim of a potential conflict. All class counsel swore that they have no affiliations with the specific research centers. Class counsel repeated that attestation at the final settlement approval hearing and added that they sit on no boards for any of the proposed recipients. As one class counsel put it, “I simply got my law degree [at Harvard], and that‘s simply the end of it.”8 The recipients are well-recognized centers focusing on the Internet and data privacy, and the district court conducted a “careful[] review” of the recipients’ “detailed proposals” and found a “substantial nexus” between the recipients and the interests of the class members.9 No one suggested that any of the centers
As an overarching matter, nothing in this record “raise[s] substantial questions about whether the selection of the recipient was made on the merits.” See Principles of the Law of Aggregate Litig. § 3.07 cmt. b. We do not suggest, however, that a party‘s prior relationship with a cy pres recipient could not be a stumbling block to approval of a settlement. Cf. Marek, 134 S.Ct. at 9 (mem.) (statement of Roberts, C.J., respecting the denial of certiorari) (recognizing that given the “fundamental concerns surrounding” cy pres awards and their increasing prevalence, the Court “may need to clarify the limits on the use of such remedies” in the future). We hold merely that, under the circumstances here, the district court did not abuse its discretion in approving the cy pres recipients.
III. Attorneys’ Fees
Turning to the issue of attorneys’ fees, the district court did not abuse its discre-
tion by approving $2.125 million in fees and $21,643.16 in costs. As an initial matter, there is no support for Objectors’ view that the settlement should have been valued at a lower amount for the purposes of calculating attorneys’ fees simply because it was cy pres-only. See generally Lane, 696 F.3d at 818 (acknowledging a 25% fee award that also involved a cy pres-only settlement). Rather, the question is whether the amount of attorneys’ fees was reasonable. In re Bluetooth Headset Prod. Liab. Litig., 654 F.3d 935, 941 (9th Cir. 2011).
In a settlement that produces a common fund for the benefit of the entire class, a court has discretion to employ either the “percentage-of-recovery” method or the “lodestar” method to calculate appropriate attorneys’ fees, so long as its discretion is exercised so as to achieve a reasonable result. See id. at 942. Here, the district court found that the requested fees were appropriate under either metric.
Under the percentage-of-recovery method, the requested fee was equal to 25% of the settlement fund. According to the district court, this percentage was commensurate with the risk posed by the action and the time and skill required to secure a successful result for the class, given that class counsel faced three motions to dismiss and participated in extensive settlement negotiations. The district court also found that this percentage hewed closely to that awarded in similar
Although not required to do so, the district court took an extra step, cross-checking this result by using the lodestar method. See In re Bluetooth, 654 F.3d at 941-44 (checking the district court‘s percentage-of-recovery fees calculation against the lodestar method, which is “calculated by multiplying the number of hours the prevailing party reasonably expended on the litigation ... by a reasonable hourly rate for the region and the experience of the lawyer“). The district court found that class counsel provided sufficient support for its lodestar calculation that fees totaled $2,126,517.25.
AFFIRMED.
WALLACE, Circuit Judge, concurring in part and dissenting in part:
I concur in Sections I and III of the majority opinion. I agree that a cy pres-only settlement was appropriate in this case and do not contend that the district court abused its discretion in calculating class counsel‘s fees.
I dissent, however, from Section II of the opinion, in which the majority blesses the district court‘s approval of the settlement, despite the preexisting relationships between class counsel and the cy pres recipients. To me, the fact alone that 47% of the settlement fund is being donated to the alma maters of class counsel raises an issue which, in fairness, the district court should have pursued further in a case such as this. The district court made no serious inquiry to alleviate that concern. Accordingly, I would vacate the district court‘s approval of the class settlement, and re-mand with instructions to hold an evidentiary hearing, examine class counsel under oath, and determine whether class counsel‘s prior affiliation with the cy pres recipients played any role in their selection as beneficiaries.
I.
As the majority opinion outlines, plaintiffs in this case alleged that Google violated class members’ privacy rights by disclosing personal information (such as search terms) to unauthorized third parties. Google‘s practice allegedly violated the federal
The settlement contained the following key terms: (1) Google agreed to pay $8.5 million into a settlement fund; (2) Google would provide notice of the settlement on its website; (3) each class representative would receive $5,000, claims administration costs would be $1 million, and attorney‘s fees would be $2.125 million (25% of the settlement fund); and (4) the remainder of the settlement fund (about $5 million) would go to six cy pres recipients. The six cy pres recipients were to be Carnegie Mellon University (21% of the remainder), the World Privacy Forum (17%), Chicago-Kent College of Law Center for Information, Society and Policy (16%), the Stanford Center for Internet and Society (16%), the Berkman Center for Internet and Society at Harvard University (15%), and the AARP Foundation (15%).
II.
We review a district court‘s approval of a class action settlement for an abuse of
As stated above, three of the cy pres distribution payments in our case are to Chicago-Kent College of Law (16%), Stanford (16%), and Harvard (15%). Attorneys for the class attended all three of these institutions. We, along with other courts and observers, have pointed out the unseemly occurrence of cy pres funds being doled out to interested parties’ alma maters. See, e.g., Nachshin v. AOL, LLC, 663 F.3d 1034, 1039 (9th Cir. 2011); Securities & Exchange Comm‘n v. Bear, Stearns & Co., Inc., 626 F.Supp.2d 402, 414-16 (S.D.N.Y. 2009); Adam Liptak, Doling out Other People‘s Money, N.Y. Times, Nov. 26, 2007 (“Lawyers and judges have grown used to controlling these pots of money, and they enjoy distributing them to favored charities, alma maters and the like“).
In response to this all-too-common development, the American Law Institute has set forth, in its Principles of the Law of Aggregate Litigation, that “[a] cy pres remedy should not be ordered if the court or any party has any significant prior affiliation with the intended recipient that would raise substantial questions about whether the selection of the recipient was made on the merits.” Principles of the Law of Aggregate Litigation § 3.07 comment b (2010) (emphasis added). Although the majority tells us correctly that no circuit has adopted the specific “prior affiliation” language, circuits have endorsed § 3.07‘s guidance regarding scrutinizing cy pres disbursements. See, e.g., In re BankAmerica Corp. Sec. Litig., 775 F.3d 1060, 1064-65 (8th Cir. 2015) (vacating a cy pres settlement because “class counsel and the district court entirely ignored this now-published ALI authority“); In re Baby Prods. Antitrust Litig., 708 F.3d 163, 172 (3d Cir. 2013) (quoting ALI § 3.07, comment a (2010)); In re Lupron Marketing and Sales Practices Litig., 677 F.3d 21, 33 (1st Cir. 2012) (citing to ALI § 3.07 and asserting that “[c]ourts have generally agreed with the ALI Principles“).
I conclude that our circuit should adopt the ALI‘s guidance as set forth in § 3.07. District courts should be required to scrutinize cy pres settlements when the proffered recipients of the funds have a “prior affiliation” with counsel, a party, or even the judge, especially when one of those players is a loyal alumni of a cy pres recipient. I do not mean to suggest that class counsel‘s alma mater can never be a cy pres beneficiary. Rather, I propose that the burden should be on class counsel to show through sworn testimony, in an on-the-record hearing, that the prior affiliation played no role in the negotiations, that other institutions were sincerely considered, and that the participant‘s alma mater is the proper cy pres recipient.
The majority responds to this line of argument by asserting that “here, we have nothing more than a barebones allegation that class counsel graduated from schools that house the Internet research centers that will receive funds.” The majority then salutes the district court‘s conclusion that there is “no indication that counsel‘s allegiance to a particular alma mater factored into the selection process,” and stresses that the cy pres recipients were a negotiated term, not chosen solely by alumni. In
I disagree fundamentally with this analysis. Our precedent requires that district courts “must be particularly vigilant not only for explicit collusion, but also for more subtle signs that class counsel have allowed pursuit of their own self-interests and that of certain class members to infect the negotiations.” In re Bluetooth, 654 F.3d at 947. In our case, we have a cy pres-only settlement. That alone raises a yellow flag. Furthermore, we have a class settlement before formal class certification. That raises another yellow flag. Lastly, we have almost half of the settlement fund, several million dollars, being given to class counsel‘s alma maters. To me, that raises a red flag. I am especially dubious of the inclusion of the Center for Information, Society and Policy at Chicago-Kent Law School (a law school attended by class counsel), which center appears to have inaugurated only a year before the parties herein agreed to their settlement. Even with these red and yellow flags, under the majority‘s holding, the burden is still on the objectors to prove more, despite the objectors’ lack of access to virtually any relevant evidence that would do so.
I would hold that the combination of a cy pres-only award, a pre-certification settlement, and the fact that almost half the cy pres fund is going to class counsel‘s alma maters, is sufficient to shift the burden to the proponents of the settlement to show, on a sworn record, that nothing in the acknowledged relationship was a factor in the ultimate choice. Here, the only sworn-to items in the record on this issue are boiler plate, one-line declarations from class counsel stating “I have no affiliation” with the subject institutions. While the majority asserts that the district court conducted a “careful review,” these terse declarations are the only shred of sworn-to evidence in the record. There was essentially nothing for the district court to review—carefully or not. Although there was some discussion between counsel and the district court during the hearings on the settlement, this was nothing more than unsworn lawyer talk during an oral argument.1
I still have many questions surrounding how these universities were chosen, such as: What other institutions were considered? Why were the non-alma mater institutions rejected? What relationship have counsel had with these universities? Have counsel donated funds to their alma maters in the past? Do counsel serve on any alma mater committees or boards? Do counsel‘s family members? How often do counsel visit their alma maters? There are many questions still lingering that have not been answered under oath. Here, as we have directed before, “the district court should have pressed the parties to substantiate their bald assertions with corroborating evidence.” Id. at 948.
Although I would vacate the parties’ settlement, I express no opinion on the definitive fairness of the parties’ agreement. It is not the province of appellate judges to “substitute our notions of fairness for those of the district judge.” Officers for