concurring in the judgment in part and dissenting in part.
This appeal requires us to decide whether there is any competent evidence to support the extradition court’s finding of probable cause that Kulvir Singh Barapind committed multiple murders and, if so, whether the crimes charged are of a “political character” which the extradition treaty between the United States and India protects from extradition. This, in turn, requires us to settle the standard by which we determine that question.
In my view there was competent evidence of the criminality of Barapind with respect to each of the three incidents at issue. Given my belief that probable causé also exists on all three charges, I must decide whether the political offense exception applies. Although we suggested a standard in Quinn v. Robinson, 783 F.2d 776, 809-10 (9th Cir.1986), for whether an offense is “incidental to” a political uprising and thus within the exception, we are now sitting en banc and so are free to consider whether that standard, or some other, should govern. I believe we should overrule Quinn’s elaboration of the “incidental to” prong and instead follow the approach articulated by the Supreme Court in Ornelas v. Ruiz, 161 U.S. 502, 511, 16 S.Ct. 689, 40 L.Ed. 787 (1896), by considering the “character of the foray, the mode of attack, the persons killed or captured, and the kind of property taken or destroyed.” Applying these factors, I cannot say that the extradition magistrate had “no choice” but to hold that the ordinary crimes Barapind committed against civilian non-combatants were of a political character. Id.
Accordingly, I would affirm the district court across the board.
I
The extradition magistrate certified that Barapind is extraditable for three offenses: the murder of Kulwant Kaur as charged in First Information Report (FIR) 89; the murders of Balwant Singh Sarhal, Amar Nath Kanugo, Suda Ram and Jasbir ’Singh as charged in FIR 34; and the murder of Sahab Singh, a.k.a. Sahbi, and the attempted murder of Makhan Ram as *754charged in FIR 100. In re Extradition of Singh, 170 F.Supp.2d 982 (E.D.Cal.2001).
The court accepted expert testimony that there was a civil war in the Punjab during the 1980s and 1990s, which was at its zenith in 1991. The Khalistan Commando Force (KCF) was a militant wing of the Sikh separation movement. KCF regularly assassinated Punjabi police and members of security forces. While in college Barapind, a Sikh, was an active member of the All India Sikh Student Federation, a group committed to establishing a sovereign Sikh nation of Khalistan to be. created from the Punjab state. He ber came president of the Federation for the District of Jalandhar in 1988.
India’s evidence upon which the extradition magistrate found probable cause of Barapind’s guilt in FIR 89- shows that on September 6,1992, Sohan Singh was sleeping on the roof of his residence in the village of Tarkham Majera, with his wife, Gurmail Kaur, and two of their sons, Par-amjit Singh and Kashmir Singh. ■ The third son, Karamjit Singh, and his wife, Kulwant Kaur, were sleeping in a room in the house. All three sons were pro-police and had been issued arms and ammunition. Around 2:00 a.m., four persons, one of whom Sohan Singh identified as Barapind, came onto the roof. Barapind shot and killed Kashmir Singh with an AK-47, then shot Paramjit Singh to death. When the assailants asked where the third son was, Gurmail Kaur told them he was sleeping in another room. Barapind stayed on the roof while the others shot Karamjit Singh and Kulwant Kaur to death.
The evidence in FIR 100 shows that at 7:15 p.m. on October 26, 1991, Makhan Ram and Sahab Singh were about to cross railway tracks when they encountered two individuals on a scooter. Makhan Ram identified the driver as Barapind, and the passenger as Gurdeep Singh, who was holding an AK-47. Gurdeep Singh opened fire, wounding Makhan Ram and killing Sahab .Singh. Two other individuals also allegedly participated.
Finally, the evidence in FIR 34 indicates that around 7:30 p.m. on April 26, 1992, Barapind, Gurdeep ' Singh, Harminder Singh, and another ydung man, armed with AK-47 rifles, came from the side of the road that leads from the village of Dhandwar’ to the village of Garhi Mohan Singh. Balwant Singh Sarhal, an ex-Member of the Legislative Assembly, along with Amar Nath Kanugo of the Deputy Commissioner’s Office, Jalandhar, and two constables, Suda Ram and Jasbir Singh, came from the side of village Garhi Mohan Singh in a “gypsy vehicle.” Barapind, Gurdeep Singh and Harminder Singh opened fire and shot and killed all four occupants. The assailants then took the constables’ weapons and left.
II
I agree with the majority that India’s affidavits are neither incompetent nor unreliable on any of the grounds asserted, and that there is probable cause of Bara-pind’s guilt on all three FIRs for reasons the majority explains. I part company on how we should treat the issue of whether Barapind’s offenses are of a “political character.” Barapind argues that Article VI of the extradition treaty with India bars his extradition because the offenses charged in FIRs 34 and 89 are of a political character. Article VI provides:
A fugitive criminal shall not be surrendered if the crime in respect of which his surrender is demanded is one of a political character, or if he proves that the requisition for his surrender has, in fact, been made with a view to try to punish him for a crime or offense of a political character.
*755Treaty for the Mutual Extradition of Criminals between the United States of America and Great Britain, Dec. 22, 1931, U.S.-Gr. Brit., T.S. No. 849 (1932). The question for the habeas court, and thus for us on appeal, is whether .the offenses charged are non-extraditable crimes within the terms of the treaty, that is, whether each is “of a political character.”
A
Barapind’s first contention — that the phrase “requisition for his surrender” in Article VI should be construed as referring to the entire extradition request — is easily resolved. To the extent Barapind’s concern is that India might try him for crimes other than those certified, I agree with the majority that the doctrine of spe-ciality prohibits the requesting country from prosecuting the fugitive for any offense except for those on which the United States agrees to extradite. Thus, Bara-pind may only be prosecuted for' nonpolitical offenses. To the extent his point is that the requisition clause prevents extradition if any offense for which he is charged is political, it makes little sense because construing the treaty in this way would skirt the doctrine of speciality and insulate non-political offenses from extradition. Beyond this, it is within the sole discretion of the Secretary of State to determine whether a country’s extradition request is a subterfuge for punishing the accused for a political crime. Quinn, 783 F.2d at 789 (citing In re Lincoln, 228 F. 70 (E.D.N.Y.1915), aff'd per curiam, 241 U.S. 651, 36 S.Ct. 721, 60 L.Ed. 1222 (1916)); Bain v. Wilkes, 641 F.2d 504, 513 (7th Cir.1981) (also citing Lincoln and Note, Executive Discretion in Extradition, 62 Colum. L.Rev. 1313, 1323 (1962)).
In default of his preferred reading, Bar-apind falls back to a somewhat different interpretation of “requisition for surrender” as referring to the underlying incident encompassing the offense. The effect in this case would be to insulate' Barapind from extradition on FIRs 89 and 34 because in both cases victims of the offenses included members of the Indian security forces (against whom there is evidence that the Sikh militant uprising was directed). This view fares no better. Article VI of the treaty protects “erime[s] or offense[s]” of a political character, not the broader incidents in which those crimes or offenses occur. If Barapind were correct, a fugitive would be protected from extradition for all crimes that occur during an attack so long as the attack had political overtones. In this way the use of lethal violence against civilians who are innocent parties, for example, would be protected even though it is the common crime of murder. I do not read the treaty as permitting any such thing.
B
The more difficult issue is what the treaty (and other treaties like it) mean by the exception for offenses of a “political character.” The treaty itself affords no help because it doesn’t define “political.”
Our court wrote extensively on the subject in Quinn. We noted a .confusion about definitions, but observed that it is “fairly well accepted” that there are two categories of political offenses — ‘“pure” and “relative.” 783 F.2d at 793. We stated that “pure” political offenses are acts aimed directly at the government and have none of the elements of ordinary crimes, while “relative” political offenses include common crimes committed in connection with a political act or for a political motive or in a, political context. Id. at 793-94 (citations omitted), -We observed that United States courts have generally adhered to an “incidence test” with two requirements: “(1) the occurrence of an up*756rising or other violent political disturbance at the time of the charged offense, and (2) a charged offense that is ‘incidental to,’ ‘in the course of,’ or ‘in furtherance of the uprising.” Id. at 797 (internal citations omitted). We declined to embrace limitations adopted by the Seventh Circuit in Eain, including in particular how it defined “uprising” as. a struggle between organized military forces, determined the legitimacy of given political objectives, and excluded violent acts against innocent civilians from the exception. Id. at 802, 808. We concluded that the incidence test protects acts of domestic, violence in connection with a struggle, for political self-determination, but does not protect acts of international terrorism. Id. at. 806. With respect to the “incidental to” component, -we adopted a “liberal nexus standard” .under which neither proof of the potential or actual effectiveness of the actions in achieving the group’s political ends nor of the fugitive’s motive or membership in the uprising group is determinative. Id- at 809. We remarked that “[i]t is for the revolutionaries, not the courts, to determine what tactics may help further their chances of bringing down or changing the government.” Id. at 810. Thus, “there is no justification for distinguishing .- between attacks. on military and civilian targets.” Id.
The panel in this case and the district court both declined to follow Quinn’s “incidental to” analysis because Quinn held that the fugitive failed to meet the “uprising” prong, thereby making the'rest of its discussion dicta. Barapind v. Enomoto, 360 F.3d 1061, 1074 n. 2, 1075 (9th Cir.2004); In re Extradition of Singh, 170 F.Supp.2d at 998. The panel then agreed with the extradition magistrate that the political offense exception is “ ‘inapplicable to shield the knowing effort to kill or injure unarmed, uninvolved, innocent civilians who are- non-combatants in the struggle.’ ” 360 F.3d at 1075 (quoting In re Extradition of Singh, 170 F.Supp.2d at 1036 (citing Ahmad v. Wigen, 726 F.Supp. 389, 405-08 (E.D.N.Y.1989) (condemning the slaughter of innocent civilians as not worthy of protection as a political offense), aff'd, 910 F.2d 1063, 1066 (2d Cir.1990); Eain, 641 F.2d at 520-21 (observing that “the indiscriminate bombing of a civilian populace is not recognized as a protected political act even when the larger ‘political’ objective of the person who sets off the bomb may be to eliminate the. civilian population of a country”); In re Extradition of Marzook, 924 F.Supp. 565, 577 (S.D.N.Y.1996) (stating that “attacks targeted at civilians do not advance any political motive other than as terrorist acts”); In re Extradition of Demjanjuk, 612 F.Supp. 544, 570 (N.D.Ohio 1985) (noting that “[t]he civilian status of the victims is also significant because the United States does not regard the indiscriminate use of violence against civilians as a political offense”), aff'd sub nom., Demjanjuk v. Petrovsky, 776 F.2d 571 (6th Cir.1985))).
We need not decide whether Quinn’s “incidental to” discussion is dicta because we are now en banc. The question for us is instead whether we should adhere to Quinn’s standard or overrule it.
i
I believe we must overrule Quinn, because indiscriminate violence against innocent persons should not qualify for the political offense exception to extradition, even if politically motivated. Nor should the propriety of committing common crimes be left to the perpetrators’ discretion. And civilians are different from the military. Overruling Quinn would realign us with the two circuits that have addressed attacks on non-combatant civilian targets and held them to be unprotected. See Ahmad, 910 F.2d at 1066 (holding that *757an attack on a commercial bus carrying civilians is not a political offense despite political motivation); Eain, 641 F.2d at 520-21 (recognizing that the civilian status of victims is of significance in considering the political offense exception).
I believe Quinn must be overruled for the additional reason that it tries to set the parameters of a “political offdnse” for all time and all places. Suffice it to say, as Justice Denman did in the leading English case In re Castioni, [1891] 1 Q.B. 149: “I do not think it is necessary or desirable that we should attempt to put into language in the shape of an exhaustive definition exactly the whole state of things, or every state of things which might bring a particular case within the description of an offence of a political character.” Id. at 155.
Having overruled Quinn in these respects, I .would look to the Supreme Court’s seminal treatment of whether an offense falls within a “political character” exception in Ornelas. The treaty in that case, with Mexico, excepted from extradition any “crime or offense of a purely political character.” A band of armed men passed over the Rio Grande from Texas into Mexico and attacked about 40 Mexican soldiers, killing and wounding some, capturing others, and taking their horses. 161 U.S. at 510, 16 S.Ct. 689. The band also violently assaulted private citizens, burning their houses, and appropriating their money, horses, and other property. Id. The fugitives’ evidence indicated that there had been a revolutionary movement on the same border against the-government the year beforehand that the purpose of their expedition was the same as the earlier one. Id. at 511, 16 S.Ct. 689, The extradition magistrate determined that the acts for which extradition was sought were not of a purely political character so as to exclude them from the treaty with Mexico; the district court disagreed; and the Supreme Court reversed based-on the character of the foray, the mode of attack, the persons killed or captured, and the kind of property taken or destroyed. Id. at 511—12, 16 S.Ct. 689.
While the terms of the treaty with Mexico — excepting crimes or offenses of a “purely” political character — are slightly narrower than the terms of the treaty with India; the Court’s approach and the factors that informed its decision are equally instructive here. The factors that the Court considered focus on specific political events and the objective acts which constitute the crimes for which extradition is sought in order to determine tvhether those crimes were part of a political revolt, insurrection, or civil war. Neither the perpetrator’s state of mind nor choice of tactics is determinative. However, it matters whether the foray is directly in aid of the uprising, how it was conducted, 'whether civilians or military were targeted, and what happened to the victims and their property.
Ornelas also tells'us that a habeas court should not disturb the extradition magistrate’s determination unless it can be said that the magistrate had “no choice” but to hold that the crime was of a political character. 161 U.S. at 511, 16 S.Ct. 689.
Considering the Ornelas factors and applying the standard of review it prescribes with respect to FIRs 34 and 89, I would uphold the district court:
FIR 31. There is ,no evidence explaining the attacks on the occupants of the gypsy vehicle. Twp of the four victims were constables or “gunmen,” which is arguably .consistent with the .KCF’s agenda of targeting Punjabi police and members of security forces. However, Balwant Singh Sarhal was ■ a former member of -the. Legislative Assembly and Amar Nath Kanugo was. then an employee in the Deputy-Commissioner’s *758Office. The expert witness could not express an opinion as to whether murder of former government ministers was an act in furtherance of the Khalistan separation movement, so I need not decide whether it would make any difference if this were the object. So far as the record discloses, none of these victims was a combatant. Barapind took the victims’ guns. For all that appears, he and his accomplices were taking advantage of a target of opportunity for mayhem, murder and theft like the marauders in Ornelas. I cannot say that the extradition magistrate had no choice but to characterize this encounter as political given that it was directed at non-combatant civilians.
FIR 89. Sohan Singh’s sons were “pro-police” collaborators who had been issued uniforms, arms and ammunition by the police for self-defense. There is evidence that they had terrorized the area by killing Khalistan militants and robbing innocent people. The police failed to take action to stop them. There is no substantial explanation in the record for why Karamjit Singh’s wife, Kulwant Kaur, was also murdered. That she was in the wrong place at the right time is one possibility, but this is discounted by the fact that Sohan Singh and his wife, Gurmail Kaur, who were on the roof with two of their sons when those sons were murdered, were left unharmed. The attack occurred at 2:00 a.m. when Barapind and his three accomplices went onto the roof to kill Paramjit and Kashmir with AK-47 rifles and then sought out Karamjit. Although the three brothers were paramilitary operatives and opponents of the Khalistan movement, Kulwant Kaur was an innocent party who was unarmed. There is no evidence that property was taken. The government does not challenge the extradition magistrate’s determination that the murders of Paramjit, Kashmir and Karamjit were linked to the movement’s political intentions, but there is nothing to suggest that the murder of Kulwant Kaur, a civilian, was committed in aid of the Khalistan separation movement. In these circumstances, I cannot say that the extradition magistrate had no choice but to hold that Kulwant Kaur’s murder was a political offense.
For these reasons, I would hold that the district court was not “palpably erroneous in law” in determining that neither the offense charged in FIR 34 nor the offense charged in FIR 89 is of a political character. Ornelas, 161 U.S. at 509, 16 S.Ct. 689.
ii
The majority takes a different tack with respect to FIRs 34 and 89, holding that the extradition court operated under “a mistaken understanding of what constitutes circuit law,” maj. op., supra at 750, or, put differently, what is dicta. In its view, the proper scope of “incidental to” was presented in Quinn, we addressed the issue and decided it in a published opinion, and it therefore became the law of the circuit even though it may not have been necessary “in some technical sense” to the disposition. I disagree that we need to go there. We are now sitting en banc, and therefore can declare the law as we believe it to be regardless of what we have previously held. This is so no matter whether a particular part of a prior opinion was necessary to its decision or not. Thus, there is no point to holding that Quinn’s “incidental to” discussion is, or is not, dicta; instead, we can, and should, decide whether its discussion is now the law of the circuit because it ought to be, and whether the district court got it right or wrong. In short, the discussion about dicta is dicta.
*759iii
In any event, I would not fault the district court for being mistaken in its understanding of dicta. It stated that the court in Quinn did not have to reach or apply the “incidental to” component of the two-part incidence test because Quinn failed to satisfy the “uprising” prong. In re Extradition of Singh, 170 F.Supp.2d at 998. We said as much ourselves in McMullen v. INS, 788 F.2d 591, 596, 598 (9th Cir.1986) (calling this part of the Quinn discussion dicta). We have also described discussions that are unnecessary to a decision as dicta. See, e.g., United States v. Johnson, 256 F.3d 895, 920 (9th Cir.2001) (en banc) (Tashima, J., concurring); Export Group v. Reef Indus., Inc., 54 F.3d 1466, 1471-72 (9th Cir.1995). So has the United States Supreme Court, see, e.g., NLRB v. Int’l Bhd. of Elec. Workers, Local 340, 481 U.S. 573, 591 n. 15, 107 S.Ct. 2002, 95 L.Ed.2d 557 (1987) (declaring that a statement in a previous decision was dictum because it “was unnecessary to the disposition”); Local 144 Nursing Home Pension Fund v. Demisay, 508 U.S. 581, 592 n. 5, 113 S.Ct. 2252, 124 L.Ed.2d 522 (1993) (declaring statements in earlier cases dicta because they were “uninvited, unargued, and unnecessary to the Court’s holdings”), and so does Black’s Law Dictionary, which defines “obiter dictum” as: “A judicial comment made while delivering a judicial opinion, but one that is unnecessary to the decision in the case and therefore not prec-edential.” Black’s Law Dictionary 1102 (8th ed.2004).
iv
If pressed to take a position, which the majority’s opinion unfortunately forces me to do, I would stick with the traditional •understanding of dictum as a statement that is not necessary to the decision.1 Like obscenity, it doesn’t seem fruitful to try to pin down a more- precise definition. I would leave it to panels, and to district courts, to sort out the occasional gratuitous observation from an authoritative holding, as we always have. They’ll know it when they see it — and if they see it differently from the active judges on this court, the remedy is a rehearing en banc which vacates the panel opinion and affords the court as a whole the opportunity to validate a prior statement or to void it. This is far more benign than the majority’s approach, which invites overwriting that may be difficult or impossible to cure.
It is one thing for a court of last resort to announce that whatever it says in a published opinion is binding, for a court of last resort regularly sits en banc, has ultimate responsibility for the efficient administration of justice within its province, and may not have enough cases to flesh out the rule being articulated. It is another for an intermediate court such as ours to make every reasoned discussion in a published opinion binding whether it is necessary or not. We speak through panels of three, and as Article III judges have authority only to decide cases and controversies. Everything that ends up in F.3d cannot possibly be the law of the circuit. Views of two or three judges in an opinion on matters that are not necessarily dispositive of the' case are no different from the same views expressed in a law review article; neither should be treated as a judicial act that is entitled to binding effect.
Accordingly, I dissent from the majority’s holding in this respect as well, for I see no reason to discuss dicta at all, let alone venture beyond traditional notions of *760what it is, and many reasons not to. I also dissent from the majority’s reversal as to FIR 34. However, I agree with its bottom line on FIR 89, and so concur in that part of the judgment.
. As Judge Posner points out, there are numerous ways to define dictum, and various reasons why holdings aré distinguished from dicta that may shed light on how a particular passage should be treated. United States v. Crawley, 837 F.2d 291 (7th Cir.1988).