Let's run through the different categories here, and see which are well-established and which are new.

  Category #1 above is well-established, and the cases the Warshak court cites are mostly about that category. For example, the Warshak court relies heavily on United States v. Phibbs, 999 F.2d 1053, 1077 (6th Cir.1993), a case involving administrative subpoenas served on a third party businesses in a criminal case. A defendant named Rojas had been indicted for drug crimes, and the DEA used administrative subpoenas to compel Rojas' credit card and telephone records. Rojas objected that the subpoenas were improper, but the court held that Rojas could not object because he didn't have a reasonable expectation of privacy in the records and this lacked standing to challenge the subpoena. This holding was unremarkable and obviously right. That was the underlying principle of United States v. Miller: if the government subpoenas your information and you have no reasonable expectation of privacy in that information, you have no standing to object to the subpoena. Thus the subpoena is covered by the usual reasonableness standard.

  What's remarkable about Warshak is that the panel seems to simply assume Category #2 and the line between Category #2 and #3. Here's what the court says about Category #2:

The government's compelled disclosure argument, while relevant, therefore begs the critical question of whether an e-mail user maintains a reasonable expectation of privacy in his e-mails vis-a-vis the party who is subject to compelled disclosure-in this instance, the ISPs. If he does not, as in Phibbs or Miller, then the government must meet only the reasonableness standard applicable to compelled disclosures to obtain the material. If, on the other hand, the e-mail user does maintain a reasonable expectation of privacy in the content of the e-mails with respect to the ISP, then the Fourth Amendment's probable cause standard controls the e-mail seizure.

  This is certainly a possible rule. It's what the lower court assumed in the Miller decision that the Supreme Court overruled. But why the standard should be probable cause isn't exactly obvious, and there are a bunch of precedents that point in the opposite direction and say the standard should be reasonableness. See, e.g., In re John Doe Proceeding, 680 N.W.2d 792 (Wis. 2004); United States v. Barr, 605 F. Supp. 114 (S.D.N.Y. 1985). These cases treat subpoenas as subpoenas, and don't automatically jump the standard up to probable cause just because a third party has a reasonable expectation of privacy in the information sought.

  Perhaps the Warshak court has a good basis for its rule — I can think of some — but it would be nice if the court had settled this very important point by explaining the issue and bothering to discuss the contrary authorities rather than just stating the conclusion.

  A few pages later, in an apparent aside in an unrelated section of the opinion, the court then carves out Category #3:

A warrant based on probable cause would not have been necessary had the government subpoenaed Warshak or given him prior notice of its intent to seek an SCA order, because the need for this higher showing would be offset by his ability to obtain judicial review before producing any e-mails. See Phibbs, 999 F.2d at 1077 (“The subpoena has to be ‘sufficiently limited in scope, relevant in purpose, and specific in directive so that compliance [would] not be unreasonable.’ If it is a subpoena duces tecum, the government does not have to secure a judicial warrant before service is effectuated. Nonetheless, ‘the subpoenaed party [must be able to] obtain judicial review of the reasonableness of the demand prior to suffering penalties for refusing to comply.'"(citing See v. City of Seattle, 387 U.S. 541, 544 (1967)).

  Woah, here, stop the presses. Why should notice "offset" the need for probable cause? If the standard of judicial review is mere reasonableness, the user presumably is going to lose every time. In the context of subpoenas, reasonableness traditionally just means that the subpoena is relevant to an investigation (even if it just goes to show no crime has been committed). So in cases where the government can provide notice, e-mail presumably is protected only very weakly: the government can provide you with notice that it is getting your e-mails just to make sure you're not committing any crimes. (At least that would be the traditional approach; I'm assuming the Warshak court doesn't also have a new theory of reasonableness for subpoenas, although I wouldn't be surprised if they had that, too.)

  Why did the Warshak court think that notice offsets the need for probable cause? Let's unpack the cases a bit to see how much law the court is making (and with how little basis in preexisting caselaw). The Warshak court cites and quotes Phibbs, which in cited See v. City of Seattle. See v. City of Seattle was case on the Fourth Amendment requirements of administrative subpoenas, which are subpoenas issued by an administrative agency rather than a court. The See Court stated:

It is now settled that, when an administrative agency subpoenas corporate books or records, the Fourth Amendment requires that the subpoena be sufficiently limited in scope, relevant in purpose, and specific in directive so that compliance will not be unreasonably burdensome.

The Phibbs case dealt with both administrative subpoenas and grand jury subpoenas (issued, at least as a matter of form, in the name of the grand jury acting through the clerk of the court). In a general section on the law of administrative subpoenas and grand jury subpoenas, the Phibbs court wrote:

Recipients of administrative subpoenas, such as those issued in accordance with 21 U.S.C. § 876, are afforded certain protections under the Fourth Amendment to the United States Constitution. The subpoena has to be “sufficiently limited in scope, relevant in purpose, and specific in directive so that compliance [would] not be unreasonable.” See v. City of Seattle, 387 U.S. 541, 544, 87 S.Ct. 1737, 1740, 18 L.Ed.2d 943 (1967).
   If it is a subpoena duces tecum, the government does not have to secure a judicial warrant before service is effectuated. Nonetheless, “the subpoenaed party [must be able to] obtain judicial review of the reasonableness of the demand prior to suffering penalties for refusing to comply.” Id. at 545, 87 S.Ct. at 1740.

  I wonder, how does any of this support the claim that notice substitutes for probable cause? If I had to guess, the Warshak court is misunderstanding who "the subpoenaed party" is in cases involving subpoenas. The subpoenaed party is the recipient of the subpoena, which in this case would be the ISP, rather than the user who has a reasonable expectation of privacy in the information, which in this case would be the user. The basic idea, going back at least as far as the Supreme Court's 1906 opinion in Hale v. Henkel, is that the party who is burdened with production has a right to challenge the subpoena. If I'm reading Warshak correctly, the court is assuming that the Internet user is the "subpoenaed party" because the subpoena ultimately relates to him; it then reads the line from See about challenging the reasonableness of subpoenas and imagines that the rule should be that a notified user can only challenge reasonableness rather than probable cause. But that's not what See was about, as I understand it; the idea that notice substitutes for probable cause seems new.

   Perhaps the Warshak court believes that there is a rough equivalence between the two standards as matter of policy. After all, in a traditional Fourth Amendment case involving no third-party storage, the government could either use a probable cause search warrant or else serve a subpoena that then triggers a reasonableness standard. So maybe they're sort of the same? But the equivalence here doesn't work, I think. What's missing is the Fifth Amendment. The major barrier to the use of subpoenas to compel evidence from suspects is the Fifth Amendment, not the Fourth Amendment. Responding to the subpoena would constitute your compelled testimony as to the authenticity, identity, and existence of the evidence that the government is seeking against you. See United States v. Hubbell. As a result, it can't ordinarily use subpoenas to compel evidence from suspects. However, if you give your information to a third party, there are no Fifth Amendment limitations on the third-party responding to a subpoena for your stuff: Fifth Amendment rights are personal and can't be asserted vicariously under Fisher v. United States. Once you have information in the hands of a third-party, the Fifth Amendment issues go away.

  This means that there really is no equivalence between the Warshak court's Categories #2 and #3. Category #3 simply makes the privacy protection granted in the "reasonable expectation of privacy" portion of the opinion a lot weaker than you would otherwise expect whenever the government knows who the bad guy is and can therefore give notice. And it does so without a particular good reason I can identify or much of a basis in the case law for making the distinction. Perhaps there is a rationale for the three categories that Judge Martin has in mind, or perhaps he is just misreading the precedents; we don't know, as this is only one of many critical questions packed into the short decision without any siginificant discussion.