(dissenting):

I dissent from my brothers’ opinion dismissing the appeal of American Ex*285press Company (Amexco) from Judge Ryan’s order granting the defendant insurers discovery of documents prepared by nonlawyer investigators under the supervision of Amexco’s attorneys despite Amexco’s claim that the documents are protected by the work-product doctrine enunciated in Hickman v. Taylor, 329 U.S. 495, 508-514, 67 S.Ct. 385, 91 L.Ed. 451 (1947), and refusing to consider Amexco’s alternative petition for a writ of mandamus. I would consider Amexco’s petition for a writ of mandamus on its merits.

Judge Ryan’s denial of Amexco’s work-product claim on its merits rested upon the ground that only “professional attorney’s work,” as distinguished from investigation, is entitled to work-product protection.1 This holding, in my opinion, squarely presents a question of law, which I have serious doubt that Judge Ryan decided correctly, and on which district judges in this circuit have divided.

Some district judges in this circuit have held, as Judge Ryan did, that the work-product doctrine protects only the work normally performed by an attorney as distinct from that usually done by an investigator. See Burke v. United States, 32 F.R.D. 213 (E.D.N.Y. 1963) (Bartels, D. J.); Brown v. New York, N. H. & H. R. R., 17 F.R.D. 324 (S.D.N.Y. 1955) (Dawson, D. J.); Szymanski v. New York, N. H. & H. R. R., 14 F.R.D. 82 (S.D.N.Y. 1952) (Sugarman, D. J.); Bifferato v. States Marine Corp., 11 F.R.D. 44 (S.D.N.Y. 1951) (Weinfeld, D. J.). Other judges have held that the doctrine extends to investigation performed by nonlawyers under the direction of attorneys. See Synder v. United States, 20 F.R.D. 7 (E.D.N.Y. 1956) (Bruchhausen, D. J.); Slifka Fabrics v. Providence Washington Ins. Co., 19 F.R.D. 374 (S.D.N.Y. 1956) (Levet, D. J.). The conflict of decisions in the district courts of this circuit is paralleled by a similar conflict among courts of appeals, compare, e. g., Alltmont v. United States, 177 F.2d 971 (3 Cir. 1949), cert. denied, 339 U.S. 967, 70 S.Ct. 999, 94 L.Ed. 1375 (1950), holding that statements obtained by nonlawyers for the use of attorneys are work-product, with, e. g., United States v. McKay, 372 F.2d 174 (5 Cir. 1967), and among commentators. Compare, e. g., Wright, Discovery, 35 F.R.D. 39, 50-51 (1964), with, e. g., Developments in the Law — Discovery, 74 Harv.L.Rev. 940, 1031 (1961).

Whether the work-product doctrine protects investigations performed by non-lawyers under the supervision of attorneys is an' issue of great importance to the members of the trial bar of this circuit. In recent years, there has been a wealth of complex litigation in which attorneys have had to make use of much nonlegal assistance. The issue presented by this appeal is also one on which district judges in this circuit have long disagreed, and one in which the harm done by an order erroneously granting discovery cannot usually be meaningfully remedied upon appeal from a final judgment. It is understandable that this issue has not reached the Court of Appeals, as the only meaningful way for a party to obtain review of a discovery order, other than by seeking a prerogative writ or certification under 28 U.S.C. § 1292(b),2 is to disobey the order and risk a contempt citation or the *286dismissal of his action or defense. We ought not to leave the bar of this circuit without some sure guide as to their duties and rights in the preparation of complex litigation.

For all these reasons, mandamus would be a proper remedy should this court conclude that Judge Ryan applied an erroneous standard. Atlass v. Miner, 265 F.2d 312 (7 Cir. 1959), aff’d, 363 U.S. 641, 80 S.Ct. 1300, 4 L.Ed.2d 1462 (1960); see 4 Moore, Federal Practice par. 26.37[4] (2 ed. 1966). I would accordingly rehear the merits of this appeal in banc, and invite the bar associations to file briefs amicus, as we did in Spanos v. Skouras Theatres Corp., 364 F.2d 161, 168 (2 Cir.), cert. denied, 385 J.S. 987, 87 S.Ct. 597, 17 L.Ed.2d 448 (1966). Taking this course to review the important question of law presented by this appeal would not, of course, commit us to entertain writs of mandamus to review ordinary questions regarding the propriety of discovery orders.

This is an attempt to obtain interlocutory review of a discovery order in a so-called “big case.”1 American Express Company (“Amexco”) appealed from an order of Judge Ryan of the Southern District of New York directing Amexco to produce certain documents for discovery. While the appeal was pending, appellees Transamerica Insurance Company and other insurance companies (“the insurers”) moved to dismiss on the ground that the order was not appealable. We ordered that the appeal be heard with argument on the insurers’ motion; we now grant the motion to dismiss and deny Amexco’s alternative application for leave to file a petition for a writ of mandamus.

Although we are amply presented with analysis of the precise legal issues before us, because of the speed with which the appeal has been handled not all of the extremely complicated underlying facts have been made similarly clear. As best we can tell from the papers before us and other matters of record, this appeal is part of the complicated litigation growing out of the “salad-oil” scandal at Bayonne, New Jersey, in the fall of 1963. At that time, Allied Crude Vegetable Oil Refining Corp. (“Allied”) “became notoriously insolvent and then a bankrupt, [and] it was discovered that it had been the vortex of frauds and the unexplained non-existence or disappearance of huge quantities of edible oils for which warehouse receipts had been issued to others.”2 Thereafter, holders of these receipts made claims against present and former wholly-owned field warehousing subsidiaries of Amex-co, both of which are now in proceedings under the Bankruptcy Act. Various related claims have been asserted against Amexco itself, counsel inform us, for “many millions of dollars — in fact, for amounts far in excess of the net worth of Amexco.” When the claims were asserted against Amexco, the insurers were put on notice; thereafter, they substantially repudiated the insurance contracts, claiming that there had been misrepresentation and concealment in obtaining the insurance. Various legal actions followed, and the many cases in the Southern District of New York have been handled by Judge Ryan. The instant appeal thus presents but a small aspect of this massive litigation; 3 it raises only one question of discovery between appellant Amexco and the appellee insurers.

Amexco and the insurers have apparently been negotiating since at least July 1965 as to the production of documents. In January 1966, by stipulation of counsel, Judge Ryan entered an order requiring Amexco, by February 15, to produce to the insurers for inspection and copying all of the “non-privileged” documents relating to certain matters. Amexco was given until February 21 to notify defendants of those categories of documents withheld from production pursuant to “any claim of privilege.” Each document thus claimed to be privileged was to be delivered to the judge “at the *279time that judicial review of such claim of privilege is sought.”

Thereafter, Amexco produced thousands of documents for discovery, but in March served a statement claiming privilege as to others. In response, defendants moved for production of certain of the latter, which were prepared after November 19, 1963, the date on which Allied had filed for relief under the Bankruptcy Act. Amexco thereupon claimed work-product “privilege” because the papers sought were reports made by a Special Investigative Unit (“the Unit”) to counsel for Amexco, and the Unit worked solely under counsel’s direction and control.

The Unit was set up by Amexco in December 1963, following Allied’s petition in bankruptcy, in anticipation of litigation. The Unit has its own quarters at Amexco’s offices and is composed of trained investigators and accountants, mostly former Federal Bureau of Investigation special agents. The Unit is headed by two Amexco employees, who had had the title of Inspector and as such had previously investigated fraud and crime. Reports made by members of the Unit were stamped “PRIVILEGED AND CONFIDENTIAL PREPARED AT THE REQUEST OF AND FOR THE USE OF COUNSEL ONLY.”

Before Judge Ryan, the insurers gave many reasons why the documents sought had to be produced under the January 1966 order. They argued principally that the nature of the documents — which had not been established by Amexco — should control, rather than under whose auspices they were created; that the rubber-stamped legend should be given no weight; e. g., it appears on a report describing the draining of water from a tank by the reporter — hardly forensic labor; and that counsel was not identified, even though at least five law firms have represented Amexco in various phases of the dispute. In turn, Amexco’s basic claim was that the documents were obtained as a re-suit of an investigation conducted by its attorneys in preparation for litigation ; since the documents are no different from those prepared by counsel themselves after a personal investigation, by “revealing the areas of interest and importance to counsel,” the documents “reflect counsel’s thought processes and litigation preparation efforts.” An enormous number of papers and a “representative selection” thereof, supposedly to facilitate the court’s review, were delivered to court under this claim.

On January 30, 1967, Judge Ryan held that to decide the claim of work-product on its merits he would have to examine the documents. However, this was unnecessary because the claim of “privilege” was based not on content but only on authorship (counsel acting through agents). Judge Ryan pointed out that Amexco did not identify “the nature or the number of the documents, the names of the persons preparing them or their relation to the client, the dates on which they were prepared and other pertinent information.” Alternatively, since the January 1966 order, which specifically called for the production of results of “investigations,” was agreed to by Amex-co when the work of the Unit had been going on for three years to the knowledge of all concerned, Judge Ryan held that Amexco waived any work-product protection as to the Unit’s work.4 Indeed, Judge Ryan noted that “the parties agreed to exchange the results of their investigations.” Therefore, the insurers’ application for discovery was granted. A short while later, the judge granted reargument but adhered to his original order. He ruled that correspondence between counsel before the stipulation regarding discovery could not vary the terms of the stipulation and that if Amexco had considered the Unit’s work to be counsel’s work-produet, it should have so provided in the stipulation. Judge Ryan also specifically stated that “the activities of this ‘Special Unit’ do not constitute attorney’s work product.”

*280On February 27, 1967, Amexco appealed from Judge Ryan’s January 30, 1967 order requiring production.5 In April, after Judge Ryan had decided the motion for reargument, the insurers moved to dismiss the appeal because the January 1967 order was not final. Upon the consent of counsel, the appeal was expedited and the panel hearing the motion to dismiss also heard the appeal. We shall first consider the question of ap-pealability.

I.

Amexco first claims that Judge Ryan’s order is a “final decision” within the gloss put upon 28 U.S.C. § 1291 by the Supreme Court in Cohen v. Beneficial Industrial Loan Corp., 337 U.S. 541, 545-547, 69 S.Ct. 1221, 93 L.Ed. 1528 (1949).6 In that ease, the Court held that denial of a motion to compel security for costs fell “in that small class which finally determine claims of right separable from, and collateral to, rights asserted in the action, too important to be denied review and too independent of the cause itself to require that appellate consideration be deferred until the whole case is adjucated.” 337 U.S. at 546, 69 S.Ct. at 1225-1226. In giving section 1291 a “practical rather than a technical construction” — before Congress provided another limited escape valve from the finality rule by enacting 28 U.S.C. § 1292 (b) — the Court nevertheless envisioned immediate appeal in only a “small class” of cases, a term hardly descriptive of the number of cases raising discovery questions. Therefore, as was to be expected, Cohen did not change the pre1949 practice of denying interlocutory review to discovery orders. 4 Moore, Federal Practice ¶ 26.37 [1. — 1] (2d ed. 1966). Thus, before and after Cohen, orders requiring the production of documents were held not “final” and therefore not appealable. For pre-1949 cases, see, e. g., Webster Coal & Coke Co. v. Cassatt, 207 U.S. 181, 28 S.Ct. 108, 52 L.Ed. 160 (1907); Apex Hosiery Co. v. Leader, 102 F.2d 702 (3d Cir. 1939) (per curiam); Bank Line, Ltd. v. United States, 163 F.2d 133, 136 (2d Cir. 1947) (dictum). For post-1949 cases, see, e. g., Paramount Film Distributing Corp. v. Civic Center Theatre, Inc., 333 F.2d 358 (10th Cir. 1964); Robinson v. Bankers Life & Casualty Co., 226 F.2d 834 (6th Cir. 1955) (per curiam). Further support could be cited for the proposition that an order requiring the production of documents is not ordinarily appealable. We note this not to stress the obvious but because in describing what this case is not — as we do below — the emphasis on the exceptional may obscure the normal statutory requirement for appeal-ability and the sound reasons for it. Among the latter are: an appellant’s ultimate right of review upon an appeal from a final judgment in the action; the elimination of unnecessary appeals, since the complaining party may win the case or settle it; the absence of irreparable harm from the vast majority of orders requiring production of documents; the potential for harassment of litigants by nuisance appeals, and the fact that any appeal tends to delay or deter trial or settlement of a lawsuit; the burden on the reviewing court’s docket from appeals of housekeeping matters in the district courts; and the slim chance for reversal of all but the most unusual discovery orders.

The rule of nonappealability is no different when a claim of attorney’s work-product under Hickman v. Taylor, 329 U.S. 495, 67 S.Ct. 385, 91 L.Ed. 451 (1947), is made. Counsel have not cited, nor have we been able to find, a single case where an assertion of work-product, either accepted or rejected by the district court, so colored the case as to cause an appellate court to assert jurisdiction in contravention of the normal rule against *281appealability of discovery orders.7 Work-product questions do not frequently get to the circuit courts, but when they do, jurisdiction does not appear to be predicated on the ground that the discovery order was final under 28 U.S.C. § 1291. Thus, in Hickman itself, the appellants (respondents in the Supreme Court) had been adjudged in contempt for refusal to obey an order to produce statements of witnesses.8 Similarly, in Alltmont v. United States, 177 F.2d 971 (3d Cir. 1949), cert. denied, 339 U.S. 967, 70 S.Ct. 999, 94 L.Ed. 1375 (1950), there was a judgment of liability, appealable in admiralty even though interlocutory, for appellant’s failure to comply with a discovery order. In New York Central R. R. v. Carr, 251 F.2d 433 (4th Cir. 1957), and Southern Ry. v. Campbell, 309 F.2d 569 (5th Cir. 1962), the party compelled to turn over claimed work-product appealed from a judgment on the merits. See also Guilford National Bank v. Southern Ry., 297 F.2d 921 (4th Cir. 1962) (contempt); Hauger v. Chicago, R. I. & Pac. R. R., 216 F.2d 501 (7th Cir. 1954) (same); Goosman v. A. Duie Pyle, Inc., 320 F.2d 45 (4th Cir. 1963) (judgment against party refused discovery).

In analogizing its situation to Cohen, Amexco’s thrust against the authorities barring appellate review is a claim of irreparable harm should discovery be erroneously granted. We do not think that the mere possibility of erroneous application of the Hickman principle to a given set of documents raises a spectre of such dire consequences that immediate appellate review as of right must follow. For such would be the inescapable conclusion if this appeal were allowed. Work-product protection is not so absolute that disclosure can never be justified. Hickman states to the contrary, 329 U.S. at 511-514, 67 S.Ct. 385. Unlike the case of a trade se-. cret, erroneous disclosure of work-product does not make almost certain the immediate destruction of a protected property right,9 cf. Covey Oil Co. v. Continental Oil Co., 340 F.2d 993 (10th Cir.), cert. denied 380 U.S. 964, 85 S.Ct. 1110, 14 L.Ed.2d 155 (1965) (appellants not parties to litigation). The interest protected is not only qualified but intangible and difficult to relate to immediate harm. Moreover, there can be appellate review of the order below even without a sanction under Fed.R.Civ.P. 37(b) (2). While success in an appeal after trial may be unlikely,10 it nevertheless can be had, e. g., Southern Ry. v. Campbell, supra, 309 F.2d 569. It is true that if reversal is obtained, the disclosure has already been made. To the extent that trial strategy is thus uncovered, the harm occasioned resembles that suffered in any retrial because of error in the first. However, that byproduct of confining appeals to final judgments has never been regarded as calamitous. Nevertheless, it may be pointed out that in a retrial each side has still been obliged to prepare independently for the first trial, a major justification for work-product protection. There is left then the possibility of a residue of significant irreparable harm— i. e., encouragement of lazy preparation— if work-product is erroneously disclosed at the discovery stage. That possibility may exist in a particular case, although *282we regard it as most improbable in this hotly contested litigation. Moreover, there are, after all, some matters which must rest primarily in the wise discretion of experienced trial judges — and when they do, the probability of recurring error is not very great. We do not suggest that the policies in support of the work-product doctrine are unimportant; we have no hesitancy in endorsing the values it protects. See Republic Gear Co. v. Borg-Warner Corp., 381 F.2d 551 (2d Cir. June 30, 1967). But we are loath to elevate those values and those policies above many others already referred to by revising the normal federal rule on appealability.

The cases relied on by appellant to justify appealability here as a “final” order under Cohen are readily distinguishable. For example, in Eisen v. Carlisle & Jacquelin, 370 F.2d 119 (2d Cir. 1966), cert. denied, 386 U.S. 1035, 87 S.Ct. 1487, 18 L.Ed.2d 598 (May 8, 1967), the district court’s order “for all practical purposes” terminated the litigation ; in Chabot v. National Securities & Research Corp., 290 F.2d 657 (2d Cir. 1961), appellants also stood to have their action terminated if the district court’s order went unreversed. No one claims that Judge Ryan’s order here has that effect. In Covey Oil, supra and Overby v. United States Fidelity & Guaranty Co., 224 F.2d 158 (5th Cir. 1955), the appellant was not a party and therefore could not obtain review on appeal from a judgment in the main action. Similarly, in Carter Prods., Inc. v. Eversharp, Inc., 360 F.2d 868 (7th Cir. 1966), discovery was sought of a non-party witness and the underlying litigation was in a different circuit, so that appellant had no other effective way to obtain appellate review of the order denying discovery.11 See also Republic Gear Co. v. Borg-Warner Corp., supra.

The rules governing the jurisdiction of the federal courts of appeals protect not only litigants, but the courts themselves. The reasons for the rules are practical ones; they have already been referred to and need not be restated. It is because of them and the precedents already discussed that we refuse to extend the Cohen doctrine to cover an appeal by a party from a discovery order based upon assertion of a work-product immunity.

II.

When a discovery question is of extraordinary significance or there is extreme need for reversal of the district court’s mandate before the case goes to judgment, there are escape hatches from the finality rule: a certification by the district court under 28 U.S.C. § 1292(b), apparently not sought here, or an extraordinary writ. It is to the latter that we now turn. Amexco further argues that if no appeal is allowed, this court should entertain jurisdiction of its application for a writ of mandamus. The most recent case in the Supreme Court on the use of extraordinary writs is Schlagenhauf v. Holder, 379 U.S. 104, 85 S.Ct. 234, 13 L.Ed.2d 152 (1964). There the district court granted a petition requiring a defendant bus driver in a negligence action to submit to mental and physical examinations under Fed.R.Civ.P. 35(a). Defendant applied for a writ of mandamus in the Court of Appeals for the Seventh Circuit. That court denied the writ on the ground that the only issues raised by mandamus went to the district judge’s power and discretion to authorize the examination; since the district court had the power to require examination of the defendant, and *283its discretion was not “so clearly abused * * * as to make the equities of this case truly extraordinary,” the petition was denied. 321 F.2d 43, 48 (1963). In the Supreme Court, respondent argued that the court of appeals should be affirmed on the ground that mandamus was not an appropriate remedy. The Court stated (379 U.S. at 110-112, 85 S.Ct. at 238):

Here petitioner’s basic allegation was lack of power in a district court to order a mental and physical examination of a defendant. * * * [T]his issue was substantial * * *.

******

We recognize that in the ordinary situation where the sole issue presented is the district court’s determination that “good cause” has been shown for an examination, mandamus is not an appropriate remedy, absent, of course, a clear abuse of discretion. * * * Here, however, the petition was properly before the court on a substantial allegation of usurpation of power in ordering any examination of a defendant, an issue of first impression that called for the construction and application of Rule 35 in a new context. The meaning of Rule 35’s requirements of “in controversy” and “good cause” also raised issues of first impression. In our view, the Court of Appeals should have also, under these special circumstances, determined the “good cause” issue, so as to avoid piecemeal litigation and to settle new and important problems.

Thus we believe that the Court of Appeals had power to determine all of the issues presented by the petition for mandamus. * * *

This is not to say, however, that, following the setting of guidelines in this opinion, any future allegation that the District Court was in error in applying these guidelines to a particular case makes mandamus an appropriate remedy. The writ of mandamus is not to be used when “the most that could be claimed is that the district courts have erred in ruling on matters within their jurisdiction.” [Footnote omitted.]

Thus, the touchstones named by the Court of review by mandamus are usurpation of power, clear abuse of discretion and the presence of an issue of first impression. Obviously, Judge Ryan had power under Rule 34 to order Amexco to produce documents for discovery in this case. Moreover, the judge did not approach abuse of discretion when two alternative grounds for his decision were Amexco’s waiver and its failure to identify the “work-product” other than by blanket assertion of its source. As to the former, while Amexco strongly urges that there could not have been any waiver below and that the insurers did not argue it (the insurers appear to agree), we will not hold that a district judge — particularly in a closely supervised “big case” — is necessarily bound by the parties’ undisclosed interpretation of the judge’s orders before or after they are signed. As to the second basis of the decision, Amexco’s submission of a “representative” sample probably bolstered the judge in his conclusion that authorship and not content was the foundation of the claim of work-product for each and every one of a tremendous number of documents. Finally, both holdings involve application of well-known law to commonplace fact, and rested on the district judge’s appraisal of facts and exercise of discretion. Whether the order below was correct or not, these bases for it involved no question of first impression.

Despite Amexco’s lengthy analysis of whether work-product extends to work done by agents of an attorney, Judge Ryan never had to reach this interesting question. Compare Southern Ry. v. Campbell, 309 F.2d 569 (5th Cir. 1962) (statements taken by claim agents not work-product), with Alltmont v. United States, 177 F.2d 971 (3d Cir. 1949), cert. denied, 339 U.S. 967, 70 S.Ct. 999, 94 L.Ed. 1375 (1950) (Hickman protects statements taken by FBI agents); cf. Guilford National Bank v. Southern Ry., 297 F.2d 921 (4th Cir. *2841962) (not reaching issue); United States v. McKay, 372 F.2d 174 (5th Cir. 1967). In fact, if anything, Judge Ryan indicated that work done by agents for an attorney might, in some circumstances, be protected by the doctrine.12 Under such circumstances, this is hardly the case to announce the “guidance” for the district courts in this circuit that Amexco claims is necessary; reaching out to assert jurisdiction by extraordinary writ is unjustified when the issues decided would bypass that controversy. While there has been occasional use of mandamus to review discovery issues, see Hartley Pen Co. v. United States District Court, 287 F.2d 324 (9th Cir. 1961); Atlass v. Miner, 265 F.2d 312 (7th Cir. 1959), aff’d on other grounds, 363 U.S. 641, 80 S.Ct. 1300, 4 L.Ed.2d 1462 (1960), normally it is not permitted. Belships Co. v. Republic of France, 184 F.2d 119 (2d Cir. 1950) (per curiam); Bank Line, Ltd. v. United States, 163 F.2d 133 (2d Cir. 1947); Paramount Film Distributing Corp. v. Civic Center Theatre, Inc., 333 F.2d 358 (10th Cir. 1964); Pennsylvania R. R. v. Kirkpatrick, 203 F.2d 149 (3d Cir. 1953). Here, with no question of power raised and the issue claimed to be of first impression not the basis for decision, we see no good reason to allow the extraordinary writ.

A final word is appropriate. Besides the obvious possibilities for abuse in the typical case if a court of appeals were to exercise intermittent supervisory power over discovery in the district courts, we feel that the nature of the present litigation presents another compelling reason for our discretionary refusal to entertain the petition. In a large and complicated lawsuit or series of lawsuits closely related, interlocutory review of such housekeeping matters as discovery would practically preclude termination of the litigation by settlement or trial within the normal lifespan of any of the parties, attorneys or judges., In,„the recent series of electrical antitrust suits, which reached gargantuan proportions,13 interlocutory review of discovery orders was granted sparingly and in clearly distinguishable situations14 It is noteworthy that through settlement and trial, this unprecedented number of suits arising out of closely related transactions has now been effectively terminated.

Since we have decided that no appeal can be taken from Judge Ryan’s order at this stage of the litigation, it is unnecessary to deal with the issues raised as to the merits of that order. Motion to dismiss appeal granted and appeal dismissed; petition for mandamus denied.