In the agreed statement of facts and issues on the present appeal the questions have been refined as follows:

 "Was the appellant entitled before the magistrate to raise as a defence:

 (a) a contention that the byelaw was ultra vires the powers granted by s.67(1) of the Transport Act 1962;

 (b) a contention that the byelaw was unreasonable;

 (c) a contention that the administrative act that led to the byelaw being used to implement a total ban on smoking in N.S.C. trains was of so unreasonable a nature that it rendered the byelaw invalid?

 Or are these matters which can be raised only by way of proceedings for judicial review in the Divisional Court?"

    It will be convenient to consider the general jurisdictional questions before examining the merits of Mr. Boddington's particular arguments. For that purpose I will concentrate on the issues raised by the case of Bugg and the judgment of the Divisional Court in Mr. Boddington's case.

III. THE DECISION IN BUGG'S CASE

    In Bugg's case the Divisional Court considered whether it is appropriate for magistrates courts hearing criminal proceedings to decide issues regarding the validity of byelaws. The defendants in two cases had entered military protected areas. They were charged with offences under byelaws. They argued that the byelaws were invalid because the areas to which the byelaws applied were insufficiently identified. The Divisional Court allowed a defendant's appeal in one case and dismissed a prosecutor's appeal in the other case. Woolf L.J. concluded that a criminal court may decide issues concerning substantive validity but not issues of procedural validity. He stated, at p. 500D:

 "So far as procedural invalidity is concerned, the proper approach is to regard byelaws and other subordinate legislation as valid until they are set aside by the appropriate court with the jurisdiction to do so. A member of the public is required to comply with byelaws even if he believes they have a procedural defect unless and until the law is held to be invalid by a court of competent jurisdiction. If before this happens he contravenes the byelaw, he commits an offence and can be punished. Where the law is substantively invalid, the position is different. No citizen is required to comply with a law which is bad on its face. If the citizen is satisfied that that is the situation, he is entitled to ignore the law."

Since the issue before the Divisional Court was undoubtedly one of substantive validity the observations of Woolf L.J. were strictly obiter. But any observations of Woolf L.J., are entitled to great weight and Woolf L.J. is of course a great expositor of public law. And he had the advantage of sitting with Pill J., a judge with extensive Divisional Court experience.

    The reasons of Woolf L.J. can be grouped under two headings. First, there are his pragmatic reasons for thinking that a criminal court is not equipped to deal with the relevant issues. Woolf L.J. said that in cases of substantive invalidity of byelaws no evidence is required whereas in cases of procedural invalidity evidence is required. The fact that evidence is required he said, may lead to different outcomes in different courts. He said that in cases of procedural invalidity the party interested in upholding a byelaw may well not be a party to the proceedings. Secondly, Woolf L.J. relied on the developments which have taken place in judicial review over the last 25 years. The principal ground of his reasoning was that, except in "flagrant" and "outrageous" cases, a byelaw remains effective until quashed.

IV. THE CORRECTNESS OF BUGG'S CASE

    Recently in Reg. v. Wicks [1998] A.C. 92, Lord Nicholls of Birkenhead and Lord Hoffmann expressed views which called into question the correctness of Bugg's case. Reg. v. Wicks was a planning case. The defendant was charged with non compliance with an enforcement notice. He attempted to challenge the validity of the enforcement notice at a criminal trial. In the leading judgment Lord Hoffmann held that as a matter of statutory interpretation "enforcement notice" in section 179(1) of the Town and Country Planning Act 1990 means a notice issued by the authority which is formally valid and has not been set aside. Accordingly, there was no defence to the criminal charge. That was the unanimous view of the House. In these circumstances the issues raised by Bugg's case did not arise and the House expressed no final view on them. In the present case those issues do arise directly and ought to be decided. Initially there was a difficulty. Counsel for the appellant and the respondent were in agreement that the observations in Bugg's case, as well as the more far reaching observations by the Divisional Court in the present case, were wrong. It would have been undesirable for the House of Lords to decide such important issues without the benefit of full argument. Fortunately, as a result of the careful and thorough written and oral submissions of Mr. Caplan and Mr. Burnett, acting as amici curiae appointed by the Attorney-General, the House has had the benefit of argument for and against the reasoning in both cases. Moreover, there has been valuable academic discussion of the issues raised by Bugg's case: see David Feldman, "Collateral challenge and judicial review; the boundary dispute continues," [1993] P.L. 37; Carl Emery, "Public or Private Law: The Limits of Procedural Reform" [1995] P.L. 450, 455-461; Dr. Christopher Forsyth, "The Metaphysic of Nullity, Invalidity, Conceptual Reasoning and the Rule of Law," Forsyth & Hare, The Golden Metwand and the Crooked Cord: Essays on Public Law in Honour of Sir William Wade (1998), pp. 152-153; Wade and Forsyth, Administrative Law, 7th ed. (1997), pp. 321-324; Craig, Administrative Law, 3rd ed. (1994), pp. 447-466. Sir Harry Woolf's Hamlyn lecture "Protection of the Public--a New Challenge" (1987), had foreshadowed the reasoning in Bugg's case. That reasoning was criticised: J. Beatson, "Public and Private in English Administrative Law" (1987) 103 L.Q.R. 34, 59-61. I have found the discussion of the problems by academic lawyers of great assistance.

    The pragmatic reasons given by Woolf L.J. need to put in context. As Lord Hoffmann observed in Reg. v. Wicks [1998] A.C. 92, 116: "the distinction between substantive and procedural invalidity appears to cut across the distinction between grounds of invalidity which require no extrinsic evidence and those which do." An issue of substantive invalidity may involve daunting issues of fact, e.g. an issue as to unequal treatment of citizens in a pluralistic society or other forms of unreasonableness. In such a case the issues of law may also be complex. In contrast an issue of procedural invalidity of a byelaw may involve minimal evidence, e.g. simply the negative fact that an express duty to consult was breached. And the question of law may be straightforward. This aspect of the pragmatic case is not persuasive. It is true, as Woolf L.J. said, that on the evidence presented to them different magistrates courts may come to different conclusions. But this factor proves too much: it applies equally to substantive validity. In any event, although a criminal court can not quash byelaws the Divisional Court can on appeal on a case stated from a decision of magistrates give a ruling which will in practice be followed by other magistrates courts. Woolf L.J. added that the party with an interest in upholding the byelaws may not be before the court. But that is also true of cases of substantive invalidity. Moreover, in a criminal case the prosecution, backed by the resources of the state, will usually put forward the case for upholding the byelaws. I therefore regard the pragmatic case in favour of a rule that magistrates may not decide issues of procedural validity, even if the distinction can be satisfactorily drawn, as questionable.

    There is also a formidable difficulty of categorisation created by Bugg's case [1993] Q.B. 473. A distinction between substantive and procedural invalidity will often be impossible or difficult to draw. Woolf L.J. recognised that there may be cases in a grey area, e.g. cases of bad faith: p. 500F. I fear that in reality the grey area covers a far greater terrain. In Associated Provincial Picture Houses Ltd. v. Wednesbury Corporation [1948] 1 K.B. 223, 229, Lord Greene M.R. pointed out that different grounds of review "run into one another." A modern commentator has demonstrated the correctness of the proposition that grounds of judicial review have blurred edges and tend to overlap with comprehensive reference to leading cases: see Fordham, Judicial Review Handbook, 2nd ed, pp. 514-521. Thus the taking into account by a decision maker of extraneous considerations is variously treated as substantive or procedural. Moreover, even Woolf L.J. categorisation of procedural invalidity is controversial. Wade and Forsyth rightly point out that contrary to normal terminology Woolf L.J. treated procedural invalidity as being not a matter of excess or abuse of power: Wade and Forsyth, Administrative Law, 7th ed., p. 323. Categorisation is an indispensable tool in the search for rationality and coherence in law. But the process of categorisation in accordance with Bugg's case which serves to carve out of the jurisdiction of criminal courts the power to decide on some issues pertinent to the guilt of a defendant, leads to a labyrinth of paths. It is nevertheless an inevitable consequence of Bugg's case that magistrates may have to rule on the satellite issue whether a particular challenge is substantive or procedural. That may involve hearing wide-ranging arguments. Even then there may be no clear cut answer. This is a factor militating against the pragmatic case on which Woolf L.J. relied in Bugg's case.

    The problems of categorisation pose not only practical difficulties. As Lord Nicholls of Birkenhead explained in Reg. v. Wicks [1998] A.C. 92 they expose a fundamental problem. About the concluding passage in Bugg's case [1993] Q.B. 473, 500, which I have quoted, he said, at p. 108:

 "On this reasoning there is not only a boundary between the two different types of invalidity. There is also an imperative need for the boundary line to be fixed and crystal clear. There can be no room for an ambiguous grey area. On this reasoning the boundary is not merely concerned with identifying the proceedings in which, as a matter of procedure, the unlawfulness issue can best be raised. Rather, the boundary can represent the difference between committing a criminal offence and not committing a criminal offence.

  "According to this reasoning, a decision on invalidity has sharply different consequences, so far as criminality is concerned, in the two types of case. Setting aside an impugned order for procedural invalidity, as distinct from substantive invalidity, has no effect on the criminality of earliest conduct. Despite a court decision that the order was not lawfully made, the defendant is still guilty of an offence, by reason of his prior conduct.

  "Further, it would seem to follow that in the case of procedural invalidity, the defendant could be convicted even after the order is set aside as having been made unlawfully, so long as the non-compliance occurred before the order was set aside. In cases of substantive invalidity the citizen can take the risk and disobey the order. If he does so, and the order is later held to be invalid, he will be innocent of any offence. In case of procedural invalidity, the citizen is not permitted to take this risk, however clear the irregularity may be."

I regard this reasoning as unanswerable. The rule of law requires a clear distinction to be made between what is lawful and what is unlawful. The distinction put forward in Bugg's case undermines this axiom of constitutional principle.

    Now I turn to modern developments in judicial review which were the principled grounds upon which Woolf L.J. relied. The first and major factor for Woolf L.J. was the proposition that except in "flagrant" and "outrageous" cases a statutory order, such as a byelaw, remains effective until it is quashed. This is a large topic on which there are confusing and contradictory dicta. It is not possible to review the subject in detail in the context of the present case. But I cannot accept the absolute proposition in Bugg without substantial qualification. Leaving to one side the separate topic of judicial review of non-legal powers exercised by non statutory bodies, I see no reason to depart from the orthodox view that ultra vires is "the central principle of administrative law" as Wade and Forsyth, Administrative Law, 7th ed., p. 41 described it. Lord Browne-Wilkinson observed in Reg. v. Hull University Visitor, Ex parte Page [1993] A.C. 682, 701:

 "The fundamental principle [of judicial review] is that the courts will intervene to ensure that the powers of public decision-making bodies are exercised lawfully. In all cases . . . this intervention . . . is based on the proposition that such powers have been conferred on the decision-maker on the underlying assumption that the powers are to be exercised only within the jurisdiction conferred, in accordance with fair procedures and, in a Wednesbury sense. . . reasonably. If the decision-maker exercises his powers outside the jurisdiction conferred, in a manner which is procedurally irregular or is Wednesbury unreasonable, he is acting ultra vires his powers and therefore unlawfully. . . ."

This is the essential constitutional underpinning of the statute based part of our administrative law. Nevertheless, I accept the reality that an unlawful byelaw is a fact and that it may in certain circumstances have legal consequences. The best explanation that I have seen is by Dr. Forsyth who summarised the position as follows in "The Metaphysic of Nullity, Invalidity, Conceptual Reasoning and the Rule of Law," at p. 159:

 "it has been argued that unlawful administrative acts are void in law. But they clearly exist in fact and they often appear to be valid; and those unaware of their invalidity may take decisions and act on the assumption that these acts are valid. When this happens the validity of these later acts depends upon the legal powers of the second actor. The crucial issue to be determined is whether that second actor has legal power to act validly notwithstanding the invalidity of the first act. And it is determined by a analysis of the law against the background of the familiar proposition that an unlawful act is void." (Emphasis supplied.)

That seems to me a more accurate summary of the law as it has developed than the sweeping proposition in Bugg's case. And Dr. Forsyth's explanation is entirely in keeping with the analysis of the formal validity of the enforcement notice in Reg. v. Wicks which was sufficient to determine the guilt of the defendant.

    That brings me to a matter of principle and precedent. In my view the holding in Bugg is contrary to established judicial review principles establish by decisions of high authority. The general rule of procedural exclusivity judicially created in O'Reilly v. Mackman [1983] 2 A.C. 237 was at its birth recognised to be subject to exceptions, notably (but not restricted to the case) where the invalidity of the decision arises as a collateral matter in a claim for infringement of private rights. The purpose of the rule was stated to be prevention of an abuse of the process of the court, and that purpose is of prime importance in determining the reach of the general rule: compare Mercury Communications Ltd. v. Director General of Telecommunications [1996] 1 W.L.R. 48, 57E, per Lord Slynn of Hadley. Since O'Reilly v. Mackman decisions of the House of Lords have made clear that the primary focus of the rule of procedural exclusivity is situations in which an individual's sole aim was to challenge a public law act or decision. It does not apply in a civil case when an individual seeks to establish private law rights which cannot be determined without an examination of the validity of a public law decision. Nor does it apply where a defendant in a civil case simply seeks to defend himself by questioning the validity of a public law decision. These propositions are established in the context of civil cases by four decisions of the House of Lords: Roy v. Kensington Family Practitioner Committee [1992] 1 A.C. 624: Chief Adjudication Officer v. Foster [1993] A.C. 754; Wandsworth London Borough Council v. Winder [1985] A.C. 461 and in particular at pp. 509-510, per Lord Fraser of Tullybelton; Mercury Communications Ltd. v. Director General of Telecommunications [1996] 1 W.L.R. 48 and in particular at p. 57B-E, per Lord Slynn of Hadley. One would expect a defendant in a criminal case, where the liberty of the subject is at stake, to have no lesser rights. Provided that the invalidity of the byelaw is or maybe a defence to the charge a criminal case must be the paradigm of collateral or defensive challenge. And in Director of Public Prosecutions v. Hutchinson [1990] 2 A.C. 783, a criminal case, the House of Lords allowed a collateral challenge to delegated legislation. The judgment in Bugg v. Director of Public Prosecutions [1993] Q.B. 473 in effect denies the right of defensive challenge in a criminal case. In my view the observations in Bugg's case are contrary to authority and principle.

    There is, above all, another matter which strikes at the root of the decision in Bugg's case. That decision contemplates that, despite the invalidity of a byelaw and the fact that consistently with Reg. v. Wicks such invalidity may in a given case afford a defence to a charge, a magistrate court may not rule on the defence. Instead the magistrates may convict a defendant under the byelaw and punish him. That is an unacceptable consequence in a democracy based on the rule of law. It is true that Bugg's case allows the defendant to challenge the byelaw in judicial review proceedings. The defendant may, however, be out of time before he becomes aware of the existence of the byelaw. He may lack the resources to defend his interests in two courts. He may not be able to obtain legal aid for an application for leave to apply for judicial review. Leave to apply for judicial review may be refused. At a substantive hearing his scope for demanding examination of witnesses in the Divisional Court may be restricted. He may be denied a remedy on a discretionary basis. The possibility of judicial review will, therefore, in no way compensate him for the loss of the right to defend himself by a defensive challenge to the byelaw in cases where the invalidity of the byelaw might afford him with a defence to the charge. My Lords, with the utmost deference to eminent judges sitting in the Divisional Court I have to say the consequences of Bugg's case are too austere and indeed too authoritarian to be compatible with the traditions of the common law. In Eshugbayi Eleko v. Government of Nigeria [1931] A.C. 662, a habeas corpus case, Lord Atkin observed, at p. 670, that "no member of the executive can interfere with the liberty or property of a British subject except on condition that he can support the legality of his action before a court of justice." There is no reason why a defendant in a criminal trial should be in a worse position. And that seems to me to reflect the true spirit of the common law.

    There is no good reason why a defendant in a criminal case should be precluded from arguing that a byelaw is invalid where that could afford him with a defence. Sometimes his challenge may be defeated by special statutory provisions on analogy with the decision in Reg. v. Wicks [1998] A.C. 92. The defence may fail because the relevant statutory provisions are held to be directory rather than mandatory. It may be held that substantial compliance is sufficient. But, if an issue as to the procedural the validity of a byelaw is raised, the trial court must rule on it.

V. SUBSIDIARY POINTS ARISING FROM BUGG'S CASE

    For the sake of completeness I need to direct attention briefly to three subsidiary matters mentioned in Bugg's case. First Woolf L.J. quoted a passage from Lord Diplock's speech in Hoffmann--La Roche & Co. A.G. v. Secretary of State for Trade and Industry [1975] A.C. 295, 366, about the presumption that subordinate legislation is valid: see Woolf L.J. [1993] Q.B. 473, 493D-F. As Lord Hoffmann explained in Reg. v. Wicks the context of the Hoffmann-La Roche case shows that the presumption of validity is not more than an evidential matter at the interlocutory stage. There is no rule that lends validity to invalid acts. In a practical world, however, a court will usually assume that subordinate legislation, and administrative acts, are valid unless it is persuaded otherwise. Secondly, Woolf L.J. said [1993] Q.B. 473, 494 that "in the case of substantive invalidity an applicant need only show the invalidity whereas in the case of procedural invalidity there is also the need for the applicant to show that he has suffered substantial prejudice." As formulated I am unable to accept this proposition. Let me pose two cases: one a breach of a duty to consult before the making of a byelaw and the other a breach of a duty to give a hearing before making an administrative decision. In both cases that establishes the ground of review. It is true that cases could occur where it might be right in regard to an established ground of judicial review to refuse a discretionary remedy and in that respect absence of prejudice may be a relevant factor: see, for example, Ridge v. Baldwin [1964] A.C. 40 and compare Bingham L.J.'s reasons in Reg. v. Chief Constable of the Thames Valley Police, Ex parte Cotton [1990] I.R.L.R. 344, as to why denial of a remedy as a matter of discretion in such a case should be a rarity. But that is altogether different from saying that prejudice is an element that an applicant must prove to establish a ground of review. Thirdly, Woolf L.J. [1993] Q.B. 473, 493 commented on the expansion of the circumstances in which courts will intervene to quash decisions. This cannot, however, be a principled ground for carving away by judicial decision part of the jurisdiction of magistrates courts. Nor can the powers of magistrates to rule on the lawfulness of byelaws be deemed to have been frozen at some date in the past. VI. THE DIVISIONAL COURT DECISION IN THE PRESENT CASE

    It is perhaps the recognition of the difficulties inherent in the distinction drawn between substantive and procedural invalidity in Bugg's case that led Auld L.J. to extend the scope of the ruling in Bugg's case by holding that all questions of invalidity of subordinate legislation and administrative decisions should be determined only in judicial review proceedings. Auld L.J. based his decision entirely on the pragmatic grounds of the inconvenience of magistrates deciding such issues. Auld L.J. said that it "would be to beckon chaos" to permit such challenges in criminal courts. While I accept that there is force in the point that it would be convenient if all public law issues could be decided in the Divisional Court, it seems to me that Auld L.J. came to an unduly pessimistic conclusion. Moreover, he failed to take into account counter arguments. Like Lloyd L.J. in Ex parte Hutchinson and Lord Hoffmann in Reg. v. Wicks [1998] A.C. 92, 116, I am impressed with the following policy considerations put forward by a Greenham Common defendant in Ex parte Hutchinson [1988] Q.B. 384, 392:

 "Coming to London to the High Court is inconvenient and expensive. Byelaws are generally local laws which have been made for local people to do with local concerns. Magistrates' courts are local courts and there is one in every town of any size in England. The cost of proceedings in a magistrates' court are far less than in the High Court. I believe this egalitarian aspect of seeking recourse to the law in a magistrates' court to be an important sign of the availability of justice for all."

Moreover, allowing a collateral or defensive challenge "avoids a cumbrous duplicity of proceedings which could only add to the already overburdened list of applications for judicial review awaiting determination in the Divisional Court" as Lord Bridge of Harwich put it in Chief Adjudication Officer v. Foster: [1993] A.C. 754, 766-767. In any event, expediency is not a sufficient and proper basis for taking away by judicial decision part of the jurisdiction of magistrates courts to rule on issues pertinent to the guilt or innocence of defendants. Moreover, the ruling of the Divisional Court is contrary to principle and precedent which permits in civil and criminal cases a collateral or defensive challenge to subordinate legislation and administrative decisions. The result of the decision of the Divisional Court is that magistrates courts will sometimes be obliged to convict defendants and to punish them despite the fact that the invalidity of the byelaw or order on which the prosecution is based affords the defendant an answer to the charge. Subject to the qualification enunciated in Reg. v. Wicks [1998] A.C. 92 such a view of the law involves an injustice which cannot be tolerated in our criminal justice system.

    It follows that the stipendiary magistrate erred in ruling that the issues raised by Mr. Boddington were beyond his jurisdiction. It further follows that the Divisional Court erred in ruling that the issues raised by Mr. Boddington could only be determined in judicial review proceedings. Mr. Boddington was entitled at the criminal trial to challenge the relevant byelaw and the administrative decision implementing the ban on smoking. In these circumstances Mr. Boddington is now entitled to a ruling on his submissions.

VII. MR. BODDINGTON'S ARGUMENTS

    The issues raised by the underlying dispute are not difficult to determine. They do not justify elaborate exposition. Byelaw 20 can quite naturally as a matter of ordinary language be accommodated within the wide words "with respect to the smoking of tobacco in railway carriages" in section 67. In my view the byelaw is valid. That leads to the attack on the administrative decision. It is true that the administrative decision interferes with the liberty of Mr. Boddington and other smokers. On the other hand, there is a conflicting interest: N.S.C. were entitled to take the view that many passengers do not wish to be exposed to tobacco fumes even in one carriage on overcrowded trains. If N.S.C. had maintained its previous policy, which permitted some smoking on its trains, that decision would not have been vulnerable to judicial review. The decision to impose the general ban is also within the range of reasonable decisions open to a decision-maker. It follows that there is no sustainable ground on which the validity of the administrative decision can be challenged.

VIII. LEGISLATIVE REFORM

    Subject to suitable and effective safeguards to protect the individual, there is a case for legislation providing for a discretionary transfer by a criminal court of public law issues to the Divisional Court. But any such reform must confront the problem created by the fact that leave to apply for judicial review is required, and that the remedies are discretionary. Those features of judicial review procedure cannot readily be reconciled with the need to ensure justice in accordance with law to a defendant in a criminal trial. Moreover, it will be necessary to take into consideration the countervailing arguments of the type put forward by the Greenham Common defendant in Ex parte Hutchinson [1988] Q.B. 384 and to those mentioned by Lord Bridge of Harwich in Chief Adjudication Officer v. Foster [1993] A.C. 754. But, above all, it must be borne in mind that there "are grave objections to giving courts discretion to decide whether governmental action is lawful or unlawful:" Wade, Administrative Law, 6th ed. (1988), p. 354. In my view any reform must take account of such concerns.

IX. THE DISPOSAL OF THE APPEAL

    Mr. Boddington has vindicated his right to challenge the byelaw and the administrative decision of which he complained. But his defence has been rejected. I would therefore dismiss the appeal.

LORD HOFFMANN

My Lords,

    I have had the advantage of reading in draft the speeches of my noble and learned friends, Lord Irvine of Lairg L.C. and Lord Steyn. For the reasons they have given I, too, would dismiss the appeal.