Internet-Draft | MNA Framework | May 2024 |
Andersson, et al. | Expires 8 November 2024 | [Page] |
This document specifies an architectural framework for the MPLS Network Actions (MNA) technologies. MNA technologies are used to indicate actions that impact the forwarding or other processing (such as monitoring) of the packet along the Label Switched Path (LSP) of the packet and to transfer any additional data needed for these actions.¶
The document provides the foundation for the development of a common set of network actions and information elements supporting additional operational models and capabilities of MPLS networks. Some of these actions are defined in existing MPLS specifications, while others require extensions to existing specifications to meet the requirements found in "Requirements for MPLS Network Actions".¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 8 November 2024.¶
Copyright (c) 2024 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
This document specifies an architectural framework for the MPLS Network Actions (MNA) technologies. MNA technologies are used to indicate actions for Label Switched Paths (LSPs) and/or MPLS packets and to transfer data needed for these actions.¶
The document provides the foundation for the development of a common set of network actions and information elements supporting additional operational models and capabilities of MPLS networks. MNA solutions derived from this framework are intended to address the requirements found in [I-D.ietf-mpls-mna-requirements]. In addition, MNA may support actions that overlap existing MPLS functionality. This may be beneficial for numerous reasons, such as making it more efficient to combine existing functionality and new functions in the same MPLS packet.¶
MPLS forwarding actions are instructions to MPLS routers to apply additional actions when forwarding a packet. These might include load-balancing a packet given its entropy, whether or not to perform fast-reroute on a failure, and whether or not a packet has metadata relevant to the forwarding actions along the path.¶
This document generalizes the concept of MPLS "forwarding actions" into "network actions" to include any action that an MPLS router is requested to take on the packet. That includes any MPLS forwarding action, but may include other operations (such as security functions, OAM procedures, etc.) that are not directly related to forwarding of the packet. MPLS network actions are always triggered by an MNA packet but may have implications for subsequent traffic, including non-MNA packets, as discussed in Section 2.4.¶
MNA technologies may redefine the semantics of the Label, Traffic Class (TC), and Time to Live (TTL) fields in an MPLS Label Stack Entry (LSE) within a NAS.¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here. These words may also appear in this document in lower case as plain English words, absent their normative meanings.¶
Although this is an Informational document, these conventions are applied to achieve clarity in the requirements that are presented.¶
This document adopts the definitions of the following terms and abbreviations from [I-D.ietf-mpls-mna-requirements] as normative: "Network Action", "Network Action Indication (NAI)", "Ancillary Data (AD)", and "Scope".¶
In addition, this document also defines the following terms:¶
Network Action Sub-Stack (NAS): A set of related, contiguous LSEs in the MPLS label stack for carrying information related to network actions. The Label, TC, and TTL values in the LSEs in the NAS may be redefined, but the meaning of the S bit is unchanged.¶
Network Action Sub-Stack Indicator (NSI): The first LSE in the NAS contains a special label that indicates the start of the NAS.¶
Abbreviation | Meaning | Reference |
---|---|---|
AD | Ancillary Data | [I-D.ietf-mpls-mna-requirements] |
BIER | Bit Index Explicit Replication | [RFC8279] |
BoS | Bottom of Stack | [RFC6790] |
bSPL | Base Special Purpose Label | [RFC9017] |
ECMP | Equal Cost Multipath | [RFC3272] |
EL | Entropy Label | [RFC6790] |
ERLD | Entropy Readable Label Depth | [RFC8662] |
eSPL | Extended Special Purpose Label | [RFC9017] |
HBH | Hop by hop | In the MNA context, this document. |
I2E | Ingress to Egress | In the MNA context, this document. |
IGP | Interior Gateway Protocol | |
ISD | In-stack data | [I-D.ietf-mpls-mna-requirements] |
LSE | Label Stack Entry | [RFC3032] |
MNA | MPLS Network Actions | [I-D.ietf-mpls-mna-requirements] |
MSD | Maximimum SID Depth | [RFC8491] |
NAI | Network Action Indicator | [I-D.ietf-mpls-mna-requirements] |
NAS | Network Action Sub-Stack | This document |
NSI | Network Action Sub-Stack Indicator | This document |
PSD | Post-stack data | [I-D.ietf-mpls-mna-requirements] and Section 3.6 |
RLD | Readable Label Depth | This document |
SID | Segment Identifier | [RFC8402] |
SPL | Special Purpose Label | [RFC9017] |
An MNA solution specifies one or more network actions to apply to an MPLS packet. These network actions and their ancillary data may be carried in sub-stacks within the MPLS label stack and/or post-stack data. A solution must specify where in the label stack the network actions sub-stacks occur, if and how frequently they should be replicated within the label stack, and how the network action sub-stack and post-stack data are encoded.¶
It seems highly likely that some ancillary data will be needed at many points along an LSP. Replication of ancillary data throughout the label stack would be highly inefficient, as would a full rewrite of the label stack at each hop, so MNA allows encoding of network actions and ancillary data deeper in the label stack, requiring implementations to look past the first LSE. Processing of the label stack past the top of stack LSE was first introduced with the Entropy Label.[RFC6790]¶
A network action sub-stack contains:¶
Network Action Sub-Stack Indicator (NSI): The first LSE in the NAS contains a special purpose label, called the MNA label, which is used to indicate the start of a network action sub-stack.¶
Network Action Indicators (NAI): Optionally, a set of indicators that describes the set of network actions. If the set of indicators is not in the sub-stack, a solution could encode them in post-stack data. A network action is said to be present if there is an indicator in the packet that invokes the action.¶
In-Stack Data (ISD): A set of zero or more LSEs that carry ancillary data for the network actions that are present. Network action indicators are not considered ancillary data.¶
Each network action present in the network action sub-stack may have zero or more LSEs of in-stack data. The ordering of the in-stack data LSEs corresponds to the ordering of the network action indicators. The encoding of the in-stack data, if any, for a network action must be specified in the document that defines the network action. In-stack data may be referenced by multiple network actions.¶
Certain network actions may also specify that data is carried after the label stack. This is called post-stack data. The encoding of the post-stack data, if any, for a network action must be specified in the document that defines the network action. If multiple network actions are present and have post-stack data, the ordering of their post-stack data corresponds to the ordering of the network action indicators.¶
A solution must specify the order for network actions to be applied to the packet for the actions to have consistent semantics. Since there are many possible orderings, especially with bit catalogs (Section 3.5.1), the solution must provide an unambiguous specification. The precise semantics of an action are dependent on the contents of the packet, including any ancillary data, and the state of the router.¶
This document assumes that the MPLS WG will select not more than one solution for the encoding of ISD and not more than one solution for the encoding of PSD.¶
A network action may need to be processed by every node along the path, or some subset of the nodes along its path. Some of the scopes that an action may have are:¶
Hop-by-hop (HBH): Every node along the path will perform the action.¶
Ingress-to-Egress (I2E): Only the last node on the path will perform the action.¶
Select: Only specific nodes along the path will perform the action.¶
If a solution supports the select scope, it must describe how it specifies the set of nodes to perform the actions.¶
This framework does not place any constraints on the scope of, or the ancillary data for, a network action. Any network action may appear in any scope or combination of scopes, may have no ancillary data, and may require in-stack data, and/or post-stack data. Some combinations may be sub-optimal, but this framework does not place any limitations on an MNA solution. A specific MNA solution may define such constraints.¶
As described in [RFC3031], legacy devices that do not recognize the MNA label will discard the packet if the top label is the MNA label.¶
Devices that do recognize the MNA label might not implement all of the network actions that are present. A solution must specify how unrecognized network actions that are present should be handled.¶
One alternative is that an implementation should stop processing network actions when it encounters an unrecognized network action. Subsequent present network actions would not be applied. The result is dependent on the solution's order of operations.¶
Another alternative is that an implementation should drop any packet that contains any unrecognized present network actions.¶
A third alternative is that an implementation should perform all recognized present network actions, but ignore all unrecognized present network actions.¶
Other alternatives may also be possible and should be specified by the solution.¶
In some solutions, an indication may be provided in the packet or in the action as to how the forwarder should proceed if it does not recognize the action. Where an action needs to be processed at every hop, it is recommended that care be taken not to construct an LSP that traverses nodes that do not support that action. It is recognised that in some circumstances it may not be possible to construct an LSP that avoids such nodes, such as when a network is re-converging following a failure or when IPFRR [RFC5714] is taking place.¶
A node that wishes to make use of MNA and apply network actions to a packet must understand the nodes that the packet will transit, whether or not the nodes support MNA, and the network actions that are to be invoked. These capabilities are presumed to be signaled by protocols that are out-of-scope for this document and are presumed to have per-network action granularity. If a solution requires alternate signaling, it must specify that explicitly.¶
If a node does not support MNA, then it is presumed to simply ignore it.¶
Readable Label Depth (RLD) is defined as the number of LSEs, starting from the top of the stack, that a router can read in an incoming MPLS packet with no performance impact. [RFC8662] introduced Entropy Readable Label Depth (ERLD). Readable Label Depth is the same concept, but generalized and not specifically associated with the Entropy Label (EL) or MNA.¶
ERLD is not redundant with RLD because ERLD specifically specifies a value of zero if a system does not support the Entropy Label. Since a system could reasonably support MNA or other MPLS functions and needs to advertise an RLD value but not support the Entropy Label, another advertised value is required.¶
A node that pushes an NAS onto the label stack is responsible for ensuring that all nodes that are expected to process the NAS will have the entire NAS within their RLD. A node SHOULD use signaling (e.g., [RFC9088], [RFC9089]) to determine this.¶
Per [RFC8662], a node that does not support EL will advertise a value of zero for its ERLD, so advertising ERLD alone does not suffice in all cases. A node MAY advertise both ERLD and RLD and SHOULD do so if its ERLD and RLD values are different. If a node's ERLD and RLD values are the same, it MAY only advertise ERLD for efficiency reasons. If a node supports MNA but does not support EL, then it SHOULD advertise RLD.¶
RLD is advertised by an IGP MSD-Type value of (TBA) and MAY be advertised as a Node Maximum Segment Identifier (SID) Depth (MSD), Link MSD, or both.¶
An MNA node MUST use the RLD determined by selecting the first advertised non-zero value from:¶
The RLD advertised for the link.¶
The RLD advertised for the node.¶
The non-zero ERLD for the node.¶
A node's RLD is a function of its hardware capabilities and is not expected to depend on the specifics of the MNA solution.¶
Several possible ways to encode NAIs have been proposed. In this section, we summarize the proposals and some considerations for the various alternatives.¶
When network actions are carried in the MPLS label stack, then regardless of their type, they are represented by a set of LSEs termed a network action sub-stack (NAS). An NAS consists of a special label, optionally followed by LSEs that specify which network actions are to be performed on the packet and the in-stack ancillary data for each indicated network action. Different network actions may be placed together in one NAS or may be carried in different sub-stacks.¶
[I-D.ietf-mpls-mna-requirements] requires that a solution not add unnecessary LSEs to the sub-stack (Section 3.1, requirement 9). Accordingly, solutions should also make efficient use of the bits within the sub-stack (except the S-bit), as inefficient use of the bits could result in the addition of unnecessary LSEs.¶
The first LSE in a network action sub-stack contains a special label that indicates a network action sub-stack. A solution has several choices for this special label.¶
A solution may reuse an existing Base SPL (bSPL). If it elects to do so, it must explain how the usage is backward compatible, including in the case where there is ISD.¶
If an existing inactive bSPL is selected that will not be backward compatible, then it must first be retired per [RFC7274] and then reallocated.¶
A solution may select a new bSPL.¶
A solution may select a new Extended SPL (eSPL). If it elects to do so, it must address the requirement for the minimal number of LSEs.¶
A solution may allow the network operator to define the label that indicates the network action sub-stack. This creates management overhead for the network operator to coordinate the use of this label across all nodes on the path using management or signaling protocols. The user-defined label could be network-wide or LSP-specific. If a solution elects to use a user-defined label, the solution should justify this overhead.¶
In the first LSE of the network action sub-stack, only the 20 bits of Label Value and the Bottom of Stack bit are used by NSI; the TC field (3 bits) and the TTL (8 bits) are not used. This could leave 11 bits that could be used for MNA purposes.¶
If the solution elects to retain the TC and TTL fields, then the first LSE of the network action sub-stack would appear as:¶
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Label | TC |S| TTL | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Label: Label value, 20 bits TC: Traffic Class, 3 bits S: Bottom of Stack, 1 bit TTL: Time To Live¶
Further LSEs would be needed to encode NAIs. If a solution elects to retain these fields, it must address the requirement for the minimal number of LSEs.¶
If the solution elects to reuse the TC and TTL fields, then the first LSE of the network action sub-stack would appear as:¶
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Label |x x x|S|x x x x x x x x| +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Label: Label value, 20 bits x: Bit available for use in solution definition S: Bottom of Stack, 1 bit¶
The solution may use more LSEs to contain NAIs. If a solution elects to use more LSEs it must address the requirement for the minimal number of LSEs.¶
A solution must have a mechanism (such as an indication of the length of the NAS) to enable an implementation to find the end of the NAS. This must be easily processed even by implementations that do not understand the full contents of the NAS. Two options are described below, other solutions may be possible.¶
A solution may use a bit per LSE to indicate whether the NAS continues into the next LSE or not. The bit may indicate continuation by being set or by being clear. The overhead of this approach is one bit per LSE and has the advantage that it can effectively encode an arbitrarily sized NAS. This approach is efficient if the NAS is small.¶
A solution may opt to have a fixed size length field at a fixed location within the NAS. The fixed size of the length field may not be large enough to support all possible NAS contents. This approach may be more efficient if the NAS is longer but not longer than can be described by the length field.¶
Advice from one hardware designer recommends a length field as this minimizes branching in the logic.¶
A solution may choose to explicitly encode the scope of each action contained in a network action sub-stack. For example, a NAS might contain Action A (HBH), Action B (HBH), and Action C (HBH). A solution may alternately choose to have the scope encoded implicitly, based on the actions present in the network action sub-stack. For example, a NAS might contain HBH scope actions: A, B, C. This choice may have performance implications as an implementation might have to parse the network actions that are present in a network action sub-stack only to discover that there are no actions for it to perform.¶
For example, suppose that an NAS is embedded in a label stack at a depth of 6 LSEs and that the NAS contains 3 actions, each with Select scope. These actions are not applicable at the current node and should be ignored. If the scope is encoded explicitly with each action, then an implementation must parse each action. However, if the scope is encoded as part of the NAS, then an implementation need only parse the start of the NAS and need not parse individual actions.¶
Solutions need to consider the order of scoped NAIs and their associated AD within individual sub-stacks and the order of per-scope sub-stacks so that network actions and the AD can be most readily found and need not be processed by nodes that are not required to handle those actions.¶
Two options for encoding NAIs are described below, other solutions may be possible. Any solution should allow the encoding of an arbitrary number of NAIs.¶
A solution may opt to encode the set of network actions as a list of bits, sometimes known as a catalog. The solution must provide a mechanism to determine how many LSEs are devoted to the catalog when the NAIs are carried in-stack. A set bit in the catalog would indicate that the corresponding network action is present.¶
Catalogs are efficient if the number of present network actions is relatively high and if the size of the necessary catalog is small. For example, if the first 16 actions are all present, a catalog can encode this in 16 bits. However, if the number of possible actions is large, then a catalog can become inefficient. Selecting only one action that is the 256th action would require a catalog of 256 bits, which would require more than one LSE when the NAIs are carried in-stack.¶
A solution may include a bit remapping mechanism so that a given domain may optimize for its commonly used actions.¶
A solution may opt to encode the set of present network actions as a list of operation codes (opcodes). Each opcode is a fixed number of bits. The size of the opcode bounds the number of network actions that the solution can support.¶
Opcodes are efficient if there are only one or two active network actions. For example, if an opcode is 8 bits, then two active network actions could be encoded in 16 bits. However, if 16 actions are required, then opcodes would consume 128 bits. Opcodes are efficient at encoding a large number of possible actions. If only the 256th action is to be selected, that still requires 8 bits.¶
A solution may carry some NAI and AD as PSD. For ease of parsing, all AD should be co-located with its NAI.¶
If there are multiple instances of post-stack data, they should occur in the same order as their relevant network action sub-stacks and then in the same order as their relevant network actions occur within the network action sub-stacks.¶
The first nibble after the label stack has been used to convey information in certain cases [RFC4385]. A consolidated view of first nibble uses is provided in [I-D.ietf-mpls-1stnibble].¶
For example, in [RFC4928] this nibble is investigated to find out if it has the value "4" or "6". If it is not, it is assumed that the packet payload is not IPv4 or IPv6, and Equal Cost Multipath (ECMP) is not performed.¶
It should be noted that this is an inexact method. For example, an Ethernet Pseudowire without a control word might have "4" or "6" in the first nibble and thus will be ECMP'ed.¶
Nevertheless, the method is implemented and deployed, it is used today and will be for the foreseeable future.¶
The use of the first nibble for Bit Index Explicit Replication (BIER) is specified in [RFC8296]. BIER sets the first nibble to 5. The same is true for a BIER payload as for any use of the first nibble: it is not possible to conclude that the payload is BIER even if the first nibble is set to 5 because an Ethernet pseudowire without a control word might begin with a 5. However, the BIER approach meets the design goal of [RFC8296] to determine that the payload is IPv4, IPv6 or a pseudowire using a control word.¶
[RFC4385] allocates 0b0000 for the pseudowire control word and 0b0001 as the control word for the pseudowire Associated Channel Header (ACH).¶
A PSD solution should specify the contents of the first nibble, the actions to be taken for the value, and the interaction with post-stack data used concurrently by other MPLS applications.¶
For MNA to be consistent across implementations and predictable in operational environments, its semantics need to be entirely predictable. An MNA solution MUST specify a deterministic order for processing each of the Network Actions in a packet. Each network action must specify how it interacts with all other previously defined network actions. Private network actions are network actions that are not publicly documented. Private network actions MUST be included in the ordering of network actions, but the interactions of private actions with other actions are outside of the scope of this document.¶
Network actions should be defined in a document and must contain:¶
Name: The name of the network action.¶
Network Action Indicator: The bit position or opcode that indicates that the network action is active.¶
Scope: The document should specify which nodes should perform the network action as described in Section 2.1.¶
State: The document should specify if the network action can modify state in the network, and if so, the state that may be modified and its side effects.¶
Required/Optional: The document should specify whether a node is required to perform the network action.¶
In-Stack Data: The number of LSEs of in-stack data, if any, and its encoding. If this is of a variable length, then the solution must specify how an implementation can determine this length without implementing the network action.¶
Post-Stack Data: The encoding of post-stack data, if any. If this is of a variable length, then the solution must specify how an implementation can determine this length without implementing the network action.¶
A solution should create an IANA registry for network actions.¶
Network operators will need to be cognizant of which network actions are supported by which nodes and will need to ensure that this is signaled. Some solutions may require network-wide configuration to synchronize the use of the labels that indicate the start of an NAS. Solution documents must make clear what management considerations apply to the solutions they are describing. Solutions documents must describe mechanisms for performing network diagnostics in the presence of MNAs.¶
An analysis of the security of MPLS systems is provided in [RFC5920], which also notes that the MPLS forwarding plane has no built-in security mechanisms.¶
Central to the security of MPLS networks is operational security of the network; something that operators of MPLS networks are well versed in. The deployment of link-level security (e.g., [MACsec]) prevents the covert acquisition of the label stack for an attack. This is particularly important in the case of a network deploying MNA, because the MNA information may be sensitive. Thus the confidentiality and authentication achieved through the use of link-level security is particularly advantageous.¶
Some additional proposals to add encryption to the MPLS forwarding plane have been suggested [I-D.ietf-mpls-opportunistic-encrypt], but no mechanisms have been agreed upon at the time of publication of this document. [I-D.ietf-mpls-opportunistic-encrypt] offers hop-by- hop security that encrypts the label stack and is functionally equivalent to that provided by [MACsec]. Alternatively, it also offers end-to-end encryption of the MPLS payload with no cryptographic integrity protection of the MPLS label stack.¶
Particular care would be needed when introducing any end-to-end security mechanism to allow an in-stack MNA solution that needed to employ on-path modification of the MNA data, or where post-stack MNA data needed to be examined on-path.¶
A cornerstone of MPLS security is to protect the network from processing MPLS labels originated outside the network.¶
Operators have considerable experience in excluding MPLS-encoded packets at the network boundaries for example, by excluding all MPLS packets and all packets that are revealed to be carrying an MPLS packet as the payload of IP tunnels. Where such packets are accepted into an MPLS network from an untrusted third party, non-MPLS packets are immediately encapsulated in an MPLS label stack specified by the MPLS network operator and MPLS packets have additional label stack entries imported as specified by the MPLS network operator. Thus, it is difficult for an attacker to pass an MPLS-encoded packet into a network or to present any instructions to the network forwarding system.¶
Within a single well-managed domain, an adjacent domain may be considered to be trusted provided that it is sufficiently shielded from third-party traffic ingress and third-party traffic observation. In such a situation, no new security vulnerabilities are introduced by MNA.¶
In some inter-domain applications (including carrier's carrier) where a first network's MPLS traffic is encapsulated directly over a second MPLS network by simply pushing additional MPLS LSEs, the contents of the first network's payload and label stack may be visible to the forwarders in the second network. Historically this has been benign, and indeed useful for ECMP. However, if the first network's traffic has MNA information this may be exposed to MNA-capable forwarders causing unpredictable behavior or modification of the customer MPLS label stack or MPLS payload. This is an increased vulnerability introduced by MNA that SHOULD be addressed in any MNA solution.¶
Several mitigations are available to an operator:¶
a) Reject all incoming packets containing MNA information that do not come from a trusted network. Note that it may be acceptable to accept and process MNA information from a trusted network.¶
b) Fully encapsulate the inbound packet in a new additional MPLS label stack such that the forwarder finds a Bottom of Stack (BoS) bit imposed by the carrier network and only finds MNA information added by the carrier network.¶
A mitigation that we reject as unsafe is having the ingress LSR push sufficient additional labels such that any MNA information received in packets entering the network from a third-party network is made inaccessible due to it being below the RLD. This is unsafe in the presence of an overly conservative RLD value which can result in the third-party MNA information becoming visible to and acted on by an MNA forwarder in the carrier network.¶
This document requests that IANA allocate a code point from the "IGP MSD-Types" registry in the "Interior Gateway Protocol (IGP) Parameters" namespace for "Readable Label Depth", referencing this document.¶
This document is the result of work started in MPLS Open Design Team, with participation by the MPLS, PALS, and DETNET working groups.¶
The authors would like to thank Adrian Farrel for his contributions and to John Drake, Toerless Eckert, and Jie Dong for their comments.¶