| IDR Working Group | S. Hares |
| Internet-Draft | Huawei |
| Intended status: Standards Track | February 7, 2016 |
| Expires: August 10, 2016 |
An Information Model for Basic Network Policy and Filter Rules
draft-hares-idr-flowspec-combo-00.txt
BGP flow specification (RFC5575) describes the distribution of filters and actions that apply when packets are received on a router with the flow specification function turned on. If one considers the reception of the packet as an event, then BGP flow specification describes a set of minimalistic Event-Match Condition-Action policies. The initial set of policy (RFC5575 and RFC7674) for this policy includes 12 types of match filters encoded in the NLRI for two types of SAFIs (IP-only SAFI, 133; VPN SAFI, 134) for IPv4. The popularity of these flow specification filters in deployment for DoS and SDN/NFV has led to the requirement for more BGP flow specification match filters in the NLRI and more BGP flow specification actions.
This document provides rules for combining new flow specification packet ECA policies which support IPv6, L2, nvo03 and MPLS match filters, and new actions. This document also provides rules for the interaction of IDR Flow Specification policy (session ephemeral policy) with policy found in I2RS (reboot ephemeral policy), and policy found in ACLs and Policy routing (configuration policy).
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 http://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 August 10, 2016.
Copyright (c) 2016 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 (http://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 Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
Section 1 of this draft contains an introduction to BGP flow specification [RFC5575] and drafts expanding the RFC5575 state. Section 2 contains the definitions related to this draft. Section 3 provides an overview of existing and proposed flow specification policy rules decribed in terms of packet event, packet match conditions, and actions (packet forwarding or packet match). The flow specification policies reviewed include policy in RFCs ([RFC5575], [RFC7674]), IDR WG documents ([I-D.ietf-idr-flow-spec-v6], [I-D.ietf-idr-flowspec-l2vpn]), and the following proposed IDR WG documents
Section 4 describes the default precedence order for BGP flow specification policy based on Flow Specification packet events, packet match conditions, and the packet match actions; and a extended community action to be used for "ordering action". Initial validation rules requires the passing of a IPv4 route associated with the BGP Flow specification rules. Section 4 also provides proposes new rules for validating BGP Flow Specification routes based on the new technologies of BGP ROAs ([RFC6482], [RFC6483]) and BGPSEC protocol [I-D.ietf-sidr-bgpsec-protocol]. Section 5 expands this precedence order to specify how the current BGP Flow specification interacts with the following non-BGP Filter packet filter forwarding specifications:
Section 6 suggests the benefits of creating a Flow Specification version 2 with a new NRLI encoding that can allow ordering of flow specification filters and actions. Section 8 describes changes for the proposed Flow Specification Yang Module ([I-D.wu-idr-flowspec-yang-cfg].
Section 9 discusses the security considerations for all the BGP Flow Specifications.
[RFC5575] describes the dissemination of flow specification rules via groups BGP Multi-Protocol NLRIs and BGP communities. A flow specification operates on packets received in a router when the flow specification feature is configured. The flow specification specifies match conditions for filters for packets received by a router and actions to do based on a match of those filters. If one considers the reception of a packet as an event, then a BGP flow specifications can be considered a set of minimalistic Event-Match Condition-Action policies (ECA policies). This set is minimalistic because there is only one event - the reception of a packet. BGP Flow specifications are BGP policy passed between peers.
The BGP flow specification policy is specified in filters contained in the MP-BGP NLRIs and actions contained within BGP Extended communities. The BGP peer propagates the flow-specifications between domains in order to automate inter-domain coordination of traffic filtering. Two applications that are using this are: distributed denial of service attack suppression and traffic filtering in BGP/MPLS VPN service. BGP. BGP flow specifications use SAFI 133 non-VPN flow specifications, and SAFI 134 for BGP VPN flow specificatinos.
BGP Flow specification are validated based on:
Originator is specified by BGP originator path attribute or transport address of the BGP peer sending the BGP Flow specification. To support BGP flow specification, implementations are required to enforce the neighbor AS in the AS_PATH attribute is in the left-most position of AS_PATH.
+-----------------------------+
| Flow Specification (FS) |
| Policy |
+-----------------------------+
^ ^
| |
| |
+--------^-------+ +-------^-------+
| FS Rule | | FS Rule |
+----------------+ +---------------+
: :
: :
......: :.....
: :
+---------V---------+ +----V-------------+
| Rule Condition | | Rule Action |
| in BGP NLRIs | | in BGP extended |
| SAFI 133, 134 | | Communities |
+-------------------+ +------------------+
: : : : : :
.....: . :..... .....: . :.....
: : : : : :
+----V---+ +---V----+ +--V---+ +-V------++--V-----++--V---+
| Match | | match | |match | | Action || action ||action|
|Operator| |Variable| |Value | |Operator||Variable|| Value|
|*1 | | | | | |(type-) || || |
+--------+ +--------+ +------+ +--------++--------++------+
*1 match operator for Types 3-12. Match operator supports
pairs of matching operators.
Figure 1: BGP Flow Specification Policy
Match operators includes a sequence of match operations each with the form [op, value] where match can match values greater, lessthan, or equal to teh value. The sequence of match operators can be combined as logical AND or ORs.
BGP Flow specification does not indicate what happens to the flow specifications if a BGP peering session closes. [RFC5575] specifies a link to received "best-match" unicast routes, but does not provide any standard way of determining whether the flow specification sent by the BGP peer is kept after the BGP session closes. It is unclear whether BGP Flow specifications disappear when a BGP session closes (denoted as BGP session ephemeral), or disapppear when the BGP module's hardware or software reboots (reboot ephemeral), or it is kept like configuration state that survives a reboot. This document in section 5 proposes that BGP Flow Specification is by default considered BGP session ephemeral disappearing when the BGP Session closes, and processes a precedence between the different types of ephemeral state.
Why is this precedence needed?
[RFC5575] states that Flow specification takes advantage of the "ACL" feature (section 1), but it does not state how BGP Flow specification interacts with ACL features. NETCONF [RFC6241] or RESTCONF [I-D.ietf-netconf-restconf] can be used to set ACL configuration state using the [I-D.ietf-netmod-acl-model] yang data module.
[I-D.litkowski-idr-flowspec-interfaceset] proposes an action which defines that a specific ordering of BGP flow-specifications and ACLs interaction for a set of interfaces for the drop/forward actions (see section 5.2 for a review). Section 5.1 proposes a default precedence between different types of flow Specification and an action. Section 5.2 proposed an action which augments [I-D.litkowski-idr-flowspec-interfaceset] to set an alternate order of precedence of flow specification drafts.
I2RS Filter-Based RIB (FB-RIB) also specifies another way to do flow filtering per packet/frame being received ([I-D.kini-i2rs-fb-rib-info-model], [I-D.hares-i2rs-fb-rib-data-model]) using a packet filter event-match_condition-action policy (draft-hares-i2rs-pkt-eca-data-model). I2RS protocol allows a I2RS Client to talk to an I2RS Agent within a routing device ([I-D.ietf-i2rs-architecture]) to set ephemermal policy which is module ephemeral and box ephemeral. Similar to BGP flow specification, the I2RS Filter-Based RIBs focus on a minimalistic event-match_condition-action (ECA) policy with a single event - the reception of a packet/frame on by a routing device. The I2RS match_conditions examine frame/packet information (L1-L4, NV03, and SFC), and I2RS match_actions that modify packet/frame information. Figure 2 shows the structure of packet filtering ECA rules from draft-hares-i2rs-pkt-eca-data-model) used by I2RS Filter-Based RIB (FB-RIB). Note that these each rule has policy rule name, policy rule order number, and rule status.
Section 5 compares the filters and actions between BGP Flow Specification, I2RS Filter-Based RIB, Filter-RIB (aka Policy-Based Routing), and the ACL. The I2RS packet filter rules also allow the rule to be ordered and named. I2RS flow-based filters are ephemeral state [I-D.ietf-i2rs-ephemeral-state] are stored as ephemeral state which is lost upon a reboot.
+-----------+ +------------+
|Rule Group | | Rule Group |
+-----------+ +------------+
^ ^
| |
| |
+--------^-------+ +-------^-----------+
| Rule | | Rule |
+----------------+ +-------------------+
: : : :
:.................: : : :
: |.........: : :
+--V--+ +--V--+ : :
| name| |order| .........: :.....
+-----+ +-----+ : :
: :
+--------------V-------+ +--V------------+
| Rule Match condition | | Rule Action |
+----------------------+ +---------------+
: : : : : :
.....: . :..... .....: . :.....
: : : : : :
+----V---+ +---V----+ +--V---+ +-V------++--V-----++--V---+
| Match | | match | |match | | Action || action ||action|
|Operator| |variable| |Value | |Operator||Variable|| Value|
+--------+ +--------+ +------+ +--------++--------++------+
Figure 2: I2RS Filter-Based RIB Policy
[RFC5575] specifies the Traffic Action Extended Community which specifies a Terminal (T) action flag and Sampling (S) flag. The sample flag indicates that "traffic sampling and logging" [is enabled] for a set of flow specifications in a BGP packet. the details of traffic sampling and logging are not specified in this standard. Logging and sampling provide valuable information to establish the impact of BGP Flow specification in order to automatic intra-AS DoS prevention or inter-AS automation of DOS or VPN traffic filters. [RFC5575] was written before the advent of yang modules that specify operational state [I-D.ietf-netmod-opstate-reqs]. [I-D.wu-idr-flowspec-yang-cfg] proposes a BGP Flow Specification Yang Data model with BGP Flow Specification configuration, operational state for BGP Flow specifications received from peers (BGP Session Ephemeral state), and statistics on the use of filters, actions, and dropped packets. Section 7 describes how the logging and notifications for BGP Flow specifications can be added to this yang module.
[RFC5575] does not require BGP Flow specifications to be passed BGPSEC [I-D.ietf-sidr-bgpsec-protocol]. [RFC5575] states "as long as traffic filtering rules are restricted to match the corresponding unicast routing paths for relevant prefixes, the security characteristics of this protocol are equivalent to existing security properties of BGP unicast properties", and "where this is not the case, this would open the door to further denial of service attack" (section 10). [I-D.eddy-idr-flowspec-exp] suggests passing BGP Flow Specification in BGPSEC. Section 10 summarizes the security issues with the current [RFC5575] and the enhancements described in this draft, and discusses the proposed fixes that that [I-D.eddy-idr-flowspec-exp] provides.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119].
The reception of a packet is the event that causes the BGP policy to enact. By default the BGP Flow specification applies to all interfaces. This can be restricted by a BGP Flow Specification Action or policy local to a node running the BGP peer session.
The definition of a packet is not limited to a IP packet (IPv4 or IPv6) but also includes mpls packets, L2 frames (802.1Q), encapsulated packets (NVGRE or VXLAN or any other NV03 encapsulation).
The same definition of the event is utilized by the I2RS Filter-based RIBs ([I-D.kini-i2rs-fb-rib-info-model] and [I-D.hares-i2rs-fb-rib-data-model] and the Filter-Based RIBs (draft-hares-rtgwg-fb-rib-data-model), and ACL filters [I-D.ietf-netmod-acl-model].
These packet events are the standardized packet events. Additional packet events for vendors may augment these standards events.
[RFC5575] defines match conditions for IPv4 to be carried with the NLRI format for 12 types of packet match events (see figure 3), and that all filters specified must be combined by a "AND". The proposed expansions to this filter list utilizing the Flow Specification NLRI are listed in figure 4. [I-D.li-idr-flowspec-rpd] proposed a BGP Attribute which contains additional flow specification filters, and actions. Figure 5 contains the match filters from this draft.
The proposals to expand flow specification beyond [RFC5575] filter specifications include:
One filter that seems obvious is the filter for the MPLS labels. However, no proposal includes this Match filter for MPLS.
The precedence order for the match filter rules was specified in [RFC5575] and expanded in [I-D.ietf-idr-flowspec-l2vpn]. The combined precedence is shown in figure 4.
Table 1: IDR WG BGP Flow Specification Match Filter
+------+--------------------+-------------+-----------------------+
|type# | Type Name | Match | Reference |
+======+====================+=============+=======================+
| 1 | Destination Prefix | IPv4 Prefix | RFC5575 |
| | | IPv6 Prefix | ietf-idr-flow-spec-v6 |
| 2 | Source Prefix | IPv4 Prefix | RFC5575 |
| | | IPv6 Prefix | ietf-idr-flow-spec-v6 |
| 3 | IP protocol |IPv4 Protocol| RFC5575 |
| | | number | |
| 3 | Next Header |IPv6 protocol| ietf-idr-flow-spec-v6 |
| 4 | Port (source or | Port number | RFC5575 |
| | destination port) | | RFC5575 |
| 5 | Source port | Port number | RFC5575 |
| 6 | Destination port | Port number | RFC5575 |
| 7 | ICMP type | ICMP type | RFC5575 |
| 8 | ICMP code | ICMP code | RFC5575 |
| 9 | TCP Flags | 1 or 2 byte | RFC5575 |
| | | bitmask for | RFC5575 |
| | | TCP flags | |
| 10 | Packet length | # of bytes | RFC5575 |
| | (for IP packet) | | |
| 11 | DSCP | IPv4 DSCP | RFC5575 |
| | | (6 bit mask)| RFC5575 |
| 11 | Traffic class | IPv6 traffic| ietf-idr-flow-spec-v6 |
| | | (8 bit mask)| |
| 12 | IPv4 Fragment | 4 bit mask | RFC5575 |
| 13 | IPv6 Flow | 20 bit flow | ietf-idr-flow-spec-v6 |
| 14 | Ethernet type | 2 bytes |ietf-idr-flowspec-l2vpn|
| 15 | Source MAC | MAC address |ietf-idr-flowspec-l2vpn|
| 16 | Destination MAC | MAC Address |ietf-idr-flowspec-l2vpn|
| 17 | DSAP in LLC | 1 octet |ietf-idr-flowspec-l2vpn|
| 18 | SSAP in LLC | 1 octet |ietf-idr-flowspec-l2vpn|
| 19 | LLC Control field | 1 octet |ietf-idr-flowspec-l2vpn|
| 20 | SNAP | 5 octets |ietf-idr-flowspec-l2vpn|
| 21 | VLAN ID | 1 or 2 bytes|ietf-idr-flowspec-l2vpn|
| 22 | VLAN COS | 3 bit COS |ietf-idr-flowspec-l2vpn|
| 23 | Inner VLAN ID | 1 or 2 bytes|ietf-idr-flowspec-l2vpn|
| 24 | Inner VLAN COS | 1 or 2 bytes|ietf-idr-flowspec-l2vpn|
+======+====================+=============+=======================+
Figure 3
Table 2: Proposed BGP Flow Specification Match Condition Filters
+------+--------------------+-------------+-----------------------+
|type# | Type Name | Match | Reference |
+======+====================+=============+=======================+
| MF-1 | Delimiter type | 2 bytes | hao-idr-flowspec-nv03 |
| | (Encapsulation type| | |
| | VXLAN or NVGRE) | | |
| | | | |
| MF-2 | VNID | 24 bit VN | hao-idr-flowspec-nv03 |
| |(virtual network ID)| | |
| | | | |
| MF-3 | Flow ID |8 bit flow ID| hoa-idr-flowspec-nv03 |
| | (NVGRE Flow ID ) | | |
| | | | |
| MF-4 | MPLS LSP | TBD | not specified |
| |(label 20 bits, | Label stack | |
| | EXP (3 bits), S Bit| | |
| | TTL (8 bits) | | |
| | | | |
| MF-5 | Interface | TBD | not specified |
| |(Group ID, intf id) | | |
+======+====================+=============+=======================+
Figure 4
Table 3: Proposed BGP Flow Specifications Match in BGP Attribute
+------+--------------------+-------------+-----------------------+
|type# | Type Name | Match | Reference |
+======+====================+=============+=======================+
| MF-6 | Time | ?? | liang-idr-bgp-flowspec|
| | | | -time |
| MF-7 | Policy from IPv4 | ?? | li-idr-flowspec-rpd |
| | Neighbor | ?? | |
| MF-8 | Policy from IPv6 | ?? | li-idr-flowspec-rpd |
| | Neighbor | ?? | |
| MF-9 | Policy with ASpath | ?? | li-idr-flowspec-rpd |
+======+====================+=============+=======================+
Figure 5
Precedence logic for BGP Flow Specifications
(RFC5575, draft-idr-bgp-flowspec-l2vpn)
flow-rule-cmp (a,b)
{
comp1 = next_component(a);
comp2 = next_component(b);
while (comp1 || comp2) {
// component_type returns infinity on end of list
if (component_type(comp1) < component_type(comp2)) {
return A_HAS_PRECEDENCE;
}
if (component_type(comp1) > component_type(comp2)) {
return B_HAS_PRECEDENCE;
}
// IP values)
if (component_type(comp1) == IP_DESTINATION || IP_SOURCE) {
common = MIN(prefix_length(comp1),prefix_length(comp2));
cmp = prefix_compare (comp1,comp2,common);
// not equal, lowest value has precedence
// equal, longest match has precedence;
} else if (component_type (comp1) == MAC_DESTINATION ||
MAC_SOURCE) {
common = MIN(MAC_address_length(comp1),
MAC_address_length(comp2));
cmp = MAC_Address_compare(comp1,comp2,common);
//not equal, lowest value has precedence
//equal, longest match has precedence
} else {
common = MIN(component_length(comp1),
component_length(comp2));
cmp = memcmp(data(comp1), data(comp2), common);
//not equal, lowest value has precedence
//equal, longest string has precedence
}
}
}
Figure 6
[RFC5575] also defines four actions which would be carried in BGP extended communities: traffic rate (in bytes), traffic action, redirect to IPv4 VPAN, and traffic marking. Traffic action has two bits Terminal bit (T) and Sample (S) bit. If the Terminal Bit is set, the the node apply all filter rules based as defined by "AND" and precedence. If the terminal bit is clear, then the flow specification process is to stop. The Sample bit implies that the flow specification enables sampling and logging for this event.
Unfortunately, [RFC5575] was unclear about the "redirect to IP VPN action" and did not handle IPv6. [RFC7674] was written to clarify [RFC5575] by clearly specifying the 3 extended communities that "IPv4 VPN" needed to support AS 4 byte, and IPv4 address Routing Distinguishers (RDs). [I-D.ietf-idr-flow-spec-v6] was written to extend this work to IPv6 filters, and to include the IPv6 flow in the filter set as figure 5 shows.
Proposals to extend these standardized actions include:
Figure 8 shows these flow specifications.
[RFC5575] indicates that the actions specified in the document represent only the "subset of filtering actions that can be interpreted across the network". As additional standardized actions occur, the non-standard action will need to have a precedence below the standardized actions.
One the probems with adding the actions is that precedence has not been set for the actions.
Table 4: BGP Flow Specifications in RFC5575 and RFC7674
+-------+--------------------+-------------+-----------------------+
|type# | Action name | action | Reference |
+=======+====================+=============+=======================+
|0x8006 | Traffic Rate | 2 octet AS | RFC5575 |
| | (in bytes ) |4 octet float| |
| | | | |
|0x8007 | Traffic Action |6 octet bit | RFC5575 |
| |(S:Sample and log, |mask:S,T bits| |
| | T:last flowspec | | |
|0x8008 | Redirect (IP VPN) |Route Target | RFC5575 and RFC7674 |
| | (RD: 2 octet AS, |(6 octet) | |
| | 4 octet value) | | |
|0x8108 | Redirect (IP VPN) |Route Target | RFC7674 |
| | (RD: 4 octet IPv4 |(6 octet) | |
| | address, 2 byte | | |
| | value) | | |
|0x8208 | Redirect (IP VPN) |Route Target | RFC7674 |
| | (RC: 4 byte AS, | | |
| | 2 byte value ) | | |
+=======+====================+=============+=======================+
Figure 7
Table 5: Proposed Flow Specification Actions
+----- -+--------------------+-------------+-----------------------+
|type# | Action name | action | Reference |
+=======+====================+=============+=======================+
| FA1 | Traffic Rate | 2 octet AS | eddy-idr-flowspec- |
| | (in packets) |4 octet float| packet-rate |
| | | | |
| FA2 | Extended Traffic | R bit | li-idr-flowspec-rpd |
| | Extension for R | P bit | Alternate action |
| | to take additional | | procedures(this draft)|
| | Flow specifications| | |
| | from BGP Flow spec | | |
| | Policy attribute | | |
| | | | |
| FA3 | Redirect to tunnel |6 octets | hao-idr-flowspec- |
| | (tunnel in |1 bit flag | redirect-to-tunnel |
| | BGP Attribute) |(C=applies to| |
| | | copies only)| |
| | | | |
| FA4 | VLAN-action | bitmask |idr-bgp-flowspec-l2vpn |
| |(push, pop, swap) | | |
| | | | |
| FA5 | TPID Action |6 octets |idr-bgp-flowspec-l2vpn |
| | (NVGRE Flow ID ) | | |
| | | | |
| FA6 | Label Action |MPLS Tag, |liang-idr-bgp-flowspec-|
| |(push/pop/swap MPLS |TTL(1 octet) | label-01 |
| |label uses Exp flag,| S bit | |
| |TTL, Stack flag (S))| | |
| | | | |
| FA7 | Alternate NLRI | validation |eddy-idr-flowspec-exp |
| | Validation | bit mask |(some functions) |
| | (mask for support | | |
| | of RFC5755, ROA | | |
| | and bgpsec-protocol| | |
| | AS path) and L2MAC | | |
| | NRLI for IP Address| | |
| | | | |
| FA8 | for Interface set | 4 Byte AS |litkowski-idr-flowspec-|
| | filter ACL + Flow | 2 byte | interfaceset |
| | specification rules| interface | |
| | | group ID | |
+=======+====================+=============+=======================+
Note: FA8 is really a filer plus an action:
FA8-filter: Restrict processing for filters to set of interfaces
FA8-Action: Forward only if: ACL + Flow-Specification filters
suggest forwarding.
Figure 8
[RFC5575] requires BGP flow specification is not required to pass in BGPSEC [I-D.ietf-sidr-bgpsec-protocol]. [RFC5575] states "as long as traffic filtering rules are restricted to match the corresponding unicast routing paths for relevant prefixes, the security characteristics of this protocol are equivalent to existing security properties of BGP unicast properties", and "where this is not the case, this would open the door to further denial of service attack" (section 10).
[RFC5575] requires an extension of the BGP route selection procedures [RFC4271] in section 9.1.2 in order to validate the BGP flow specification NLRI. The BGP Flow Specification NLRI is valid if and only if:
This set of validation requirements also require that BGP implementations are required to enforce the AS_PATH attribute having the neighbor AS in the left-most position.
These validation steps required a unicast IPv4 or IPv6 route be transmitted with L2VPN ([I-D.ietf-idr-flowspec-l2vpn]) and the NV03 flow specifications [I-D.hao-idr-flowspec-nvo3] to validate the path. These specifications do not provide additional details on any additional validation needed for the L2VPN or NV03 Case.
Since [RFC5575] BGP Route Origin validation [RFC6482] has been standardized, and the BGPSEC protocol [I-D.ietf-sidr-bgpsec-protocol] has been developed. [I-D.eddy-idr-flowspec-exp] specifies cryptographic enhancements that include:
This document describes the precedence of these BGP security features.
BGP Flow specification is session ephemeral state which will not persist when the BGP peer session closes. I2RS Filter-Based RIB is reboot ephemeral state which will not persist when the routing entity reboots. Policy RIB (aka Filter Forwarding RIB) and ACLs are configuration state which can persist over the reboot of a system.
This precedence within BGP Session Ephemeral state depends on the preference associated with valid BGP Session flow specification NLRI received within a BGP State. Since [RFC5575] was published, additional mechanisms to validate originating prefixes with an AS with Prefix Orgin Validation (ROA), and the BGPSEC Secure Path have been standardized. The precedence of these mechanisms should be from BGP Security to ROA to [RFC5575]. The BGP peers determine that a BGP Flow specification is valid if and only if one of the following cases:
The best match is defined to be the longest-match NLRI with the highest preference.
Match conditions depends on an "AND" of all rules within a Flow Specification policy. A Flow specification policy is defined by a sequence of BGP Flow specification NLRIs with filter-match rules. The sequence of Flow Specification rules are terminate Traffic Action with a T-Bit flag set to zero.
Match condition processing occurs in the following overall precedence:
Table 6 in figure 9 shows the filter by filter precedence order. All flow specification filters combine as an "AND" of all filters. A re-ordering of match filters is only possible in the the proposed version 2 of BGP Flow specification.
Table 6: Flow Specification Match Filter Precedence Order
+------+--------------------+-------------+-----------------------+
|type# | Type Name | Match | Reference |
+======+====================+=============+=======================+
| 1 | Destination Prefix | IPv4 Prefix | RFC5575 |
| | | IPv6 Prefix | ietf-idr-flow-spec-v6 |
| 2 | Source Prefix | IPv4 Prefix | RFC5575 |
| | | IPv6 Prefix | ietf-idr-flow-spec-v6 |
| 3 | IP protocol |IPv4 Protocol| RFC5575 |
| | | number | |
| 3 | Next Header |IPv6 protocol| ietf-idr-flow-spec-v6 |
| 4 | Port (source or | Port number | RFC5575 |
| | destination port) | | RFC5575 |
| 5 | Source port | Port number | RFC5575 |
| 6 | Destination port | Port number | RFC5575 |
| 7 | ICMP type | ICMP type | RFC5575 |
| 8 | ICMP code | ICMP code | RFC5575 |
| 9 | TCP Flags | 1 or 2 byte | RFC5575 |
| | | bitmask for | RFC5575 |
| | | TCP flags | |
| 10 | Packet length | # of bytes | RFC5575 |
| | (for IP packet) | | |
| 11 | DSCP | IPv4 DSCP | RFC5575 |
| | | (6 bit mask)| RFC5575 |
| 11 | Traffic class | IPv6 traffic| ietf-idr-flow-spec-v6 |
| | | (8 bit mask)| |
| 12 | IPv4 Fragment |4 bit mask | RFC5575 |
| 13 | IPv6 Flow |20 bit flow | ietf-idr-flow-spec-v6 |
| 14 | Delimiter type | 2 bytes | hao-idr-flowspec-nv03 |
| MF-1 | (Encapsulation type| | |
| | VXLAN or NVGRE) | | |
| | | | |
| 15 | VNID | 24 bit VN | hao-idr-flowspec-nv03 |
| MF-2 |(virtual network ID)| | |
| | | | |
| 16 | Flow ID |8 bit flow ID| hoa-idr-flowspec-nv03 |
| MF-3 | (NVGRE Flow ID ) | | |
| | | | |
| 17 | Segment ID | | |
|18-25 | Other packet ids | | |
| | above MPLS | | |
| 29 | MPLS LSP | TBD | not specified |
| MF-4 |(label 20 bits, | Label stack | |
| | EXP (3 bits), S Bit| | |
| | TTL (8 bits) | | |
| | | | |
| | | | |
| 30 | Ethernet type | 2 bytes |ietf-idr-flowspec-l2vpn|
| 31 | Source MAC |MAC address |ietf-idr-flowspec-l2vpn|
| 32 | Destination MAC |MAC Address |ietf-idr-flowspec-l2vpn|
| 33 | DSAP in LLC | 1 octet |ietf-idr-flowspec-l2vpn|
| 34 | SSAP in LLC | 1 octet |ietf-idr-flowspec-l2vpn|
| 35 | Control filed in |
| | LLC| |1 octet |ietf-idr-flowspec-l2vpn|
| 36 | SNAP | 5 octet |ietf-idr-flowspec-l2vpn|
| 37 | VLAN ID |1 or 2 bytes |ietf-idr-flowspec-l2vpn|
| 38 | VLAN COS | 3 bit COS |ietf-idr-flowspec-l2vpn|
| 39 | Inner VLAN ID |1 or 2 bytes |ietf-idr-flowspec-l2vpn|
| 40 | Inner VLAN COS |1 or 2 bytes |ietf-idr-flowspec-l2vpn|
| 41 | Interface | TBD | not specified |
| |(Group ID, intf id) | | |
| 42 |Time | | |
| 65 |FCFS matches | | non-standard actions |
+======+====================+=============+=======================+
Figure 9
Some BGP Flow Specification actions can conflict with other BGP Flow specification Actions. Table 7 in figure 10 shows the default precedence order and the potential conflicting actions. Existing actions with conflicts denote the default action taken on conflicting actions.
Each flow specification that specifies a BGP action must create a "BGP Flow Specification Action Conflicts" section within the flow specification. In this section, the flow specification must point to this document indicating the precedence between actions, and indicate how the action handles the conflict. All Standards actions have precedence overall FCFS actions incoded in BGP Extended Communities.
R-Policy bit - Additional BGP Version 2 Flow specification has additional filters and policy in BGP Attribute X.
TP-Mod bit - make modifications to packet before sending to the IP-VPN via a tunnel,
R-Intf bit - process restrict to interface sets
Two bits are added to the Extended Traffic Action Flag so that the total flags are:
Table 7 - Action Precedence and Conflicts between Actions
+-----+---------------------+----------------------------------+
|order| Action | Possible Conflicting Actions |
+=====+=====================+==================================+
| FA7 | Alternate NLRI | none |
| 1 | Validation | |
| | (mask for support | |
| | of RFC5755, ROA | |
| | and bgpsec-protocol | |
| | AS path) and L2MAC | |
| | NRLI for IP Address | |
| | | |
| 2 | Traffic Rate(0x8006)| Traffic rate in packets (FA1) |
| | in bytes | |
| | | Default Conflict action: |
| | | Allow traffic monitoring by bytes|
| | | and packets, but process byte |
| | | rate limit checks first |
| | | |
| 3 | Traffic Rate (FA1) | traffic rate in bytes (0x8006) |
| | in packets | |
| | | Default Conflict action: same |
| | | as in Traffic Rate action |
| | | conflict |
| | | |
| 4 | Traffic Action | Extended Traffic action with |
| | (0x8007) | "R-Policy" bit(FA2), "TN-P" bit, |
| | | R-intf bit |
| | | |
| | | Default conflict action: Process |
| | | Traffic Action, then Extended |
| | | traffic action |
| | | |
| 5 | Extended Traffic | Traffic Action (0x8007) |
| | Action (FA2) |"R" bit(FA2), "TN-P" bit (above) |
| | | R-Intf bit |
| | | |
| | | Default conflict action: Process |
| | | Traffic action, then extended |
| | | traffic action |
| | | |
| 6 | Redirect to IP-VPN | Redirect to IP Tunnel (FA3) |
| | 0x8008: 2 byte AS RD| VLAN-action (FA4), |
| | 0x8108: 4 byte IP RD| TPID-action (FA5) |
| | 0x8208: 4 byte AS RD| Label-action (FA6) |
| | | interface set (FA7) |
| | | |
| | | Default Conflict action: |
| | | Process forward to IP-VPN first |
| | | and ignore other conflicting |
| | | actions unless TN-Mod bit set in |
| | | Extended action.
| | | If TN-Mod set then process the |
| | | conflict actions which change |
| | | the packet prior to forwarding |
| | | the packet via tunnel to IP-VPN. |
| | | |
| | | If I bit set, process interface |
| | | restriction's narraowing of scope|
| | | to certain interfaces before |
| | | processing other options, and |
| | | process interface restrictions |
| | | implied in outboudn direction |
| | | before sending packet. |
| | | outbound policy before any other |
| | | If "R" bit set use version 2 of |
| | | BGP Flow Specification handling |
| | | |
| 7 | Redirect to IP | Redirect to IP VPN (0x8008, |
| | Tunnel (FA3) | 0x8108, 0x8208) |
| | | VLAN-action (FA4), |
| | | TPID-action (FA5), |
| | | Label action (FA6), |
| | | interface set (FA7) |
| | | |
| | | Default Conflict actions: |
| | | Refer to processing in redirect |
| | | IP-VPN tunnel |
| | | |
| 8 | VLAN action (FM4) | Redirect to IP-VPN (0x8008, |
| | | 0x8108, 0x8208), |
| | | Redirect to tunnel (FA3), |
| | | VLAN-action (FA4), |
| | | TPID-action (FA5), |
| | | Label action (FA6), |
| | | interface set (FA7) |
| | | |
| | | Default Conflict actions: |
| | | Refer to processing in redirect |
| | | IP-VPN tunnel |
| | | |
| 9 | TPID action (FM5) | Redirect to IP-VPN (0x8008, |
| | | 0x8108, 0x8208), |
| | | Redirect to tunnel (FA3), |
| | | VLAN-action (FA4), |
| | | TPID-action (FA5), |
| | | Label action (FA6), |
| | | interface set (FA7) |
| | | |
| | | Default Conflict actions: |
| | | Refer to processing in redirect |
| | | IP-VPN tunnel |
| | | |
| 10 | Label Action (FM6) | Redirect to IP-VPN (0x8008, |
| | | 0x8108, 0x8208), |
| | | Redirect to tunnel (FA3), |
| | | VLAN-action (FA4), |
| | | TPID-action (FA5), |
| | | Label action (FA6), |
| | | interface set (FA7) |
| | | |
| | | Default Conflict actions: |
| | | Refer to processing in redirect |
| | | IP-VPN tunnel |
| | | |
| 11 | interface Set (FM8a)| Redirect to IP-VPN (0x8008, |
| | | 0x8108, 0x8208), |
| | | Redirect to tunnel (FA3), |
| | | VLAN-action (FA4), |
| | | TPID-action (FA5), |
| | | Label action (FA6), |
| | | |
| | | Default Conflict actions: |
| | | Refer to processing in redirect |
| | | IP-VPN tunnel |
| | | |
| 12 | Filter precedence | reorder default filter precedence|
| | (FM8b) | 0 = BGP Flow-Spec only |
| | [proposed] | 1 = ACL + BGP Flow-Spec |
| | | 2 = I2RS FB-RIB + BGP FS |
| | | 3 = ACL + I2RS FB-FIB + BGP FS |
| | | 4 = Config FB-RIB + BGP FS |
| | | 5 = ACL + config FB-RIB + BGP FS |
| | | 6 = Config FB-RIB + I2RS FB-RIB +|
| | | BGP FS |
| | | 7 = ACL + config FB-FIB + I2RS |
| | | |
|13-63| | Reserved for other standards |
| | | actions |
| | | |
|65+ | FCFS actions | FCFS Actions |
+=====+=====================+==================================+
Figure 9
[RFC5575] allowed for non-IETF standardized Flow Specification filters and extended community actions. The beginning order of precedence for non-IETF standardized FCFS BGP Flow specification match filters is 65. The network management yang modules SHOULD store the BGP Flow Specification match type byte for both IETF Standardized BGP Flow Specification Match Filters, FCFS BGP BGP Flow Specification Match filters.
[RFC7153]allows for FCFS (First Come First Serve) allocation of BGP transitive types. If an action is specified in the FCFS registry, the default precedence is after all standardized BGP Flow Specification actions(action 65+). The BGP Flow Specification Yang models should store the Extended Community value for the FCFS based Flow Specification action. If the precedence ordering has been changed by the FCFS, this should be stored in the configuration of BGP Flow Specification and in the operational state.
There are the following ways to get ordered filters and actions:
To get ordered an ordered match field in [I-D.li-idr-flowspec-rpd] the following additions would need to be made for the match field format:
match type [bit 1 - deny/permit]
0-permit, 1 -deny
+------------------------+
| match type (2 octets) |
+------------------------+
| number of sub-TLVS |
| (2 octets) |
+------------------------+
| sub-TLVs (variable) |
| +====================+ |
| | order (2 octets) | |
| +--------------------+ |
| | type (2 octets) | |
| +--------------------+ |
| | length (2 octets) | |
| +--------------------+ |
| | value (variable) | |
| +====================+ |
+------------------------+
figure 10 - match field revision
The action field would be expanded to include an action order field (2 octets) as follows:
+--------------------------+ | Action order (2 octets) | +--------------------------+ | Action type (2 octets) | +--------------------------+ | Action length (2 octets) | +--------------------------+ | Action Values (variable) | +--------------------------+ figure 11 - Action field revision
The new BGP NLRI would have the following format to order filters:
+------------------------+ |length (2 octets) | +------------------------+ | number of sub-TLVS | | (2 octets) | +------------------------+ | sub-TLVs (variable) | | +====================+ | | | order (2 octets) | | | +--------------------+ | | | type (2 octets) | | | +--------------------+ | | | length (2 octets) | | | +--------------------+ | | | value (variable) | | | +====================+ | +------------------------+ Figure 12 - NRLI revision
+--------------------------+ | order (2 octets) | +--------------------------+ | Action type (2 octets) | +--------------------------+ | Action length (2 octets) | +--------------------------+ | Action Values (variable) | +--------------------------+ Wide Community Atom figure 13
The BGP Wide community would need to have an atom (TBD) that indicates BGP Flow Specification actions. The atom would have the following information within it:
This precedence for flow policy amoung routing functions SHOULD go from the most dynamic overwriting the the least dynamic. The order from dynamic to least is:
The filtering process for a packet received should attempt to match the more dynamic policy prior to matching a less dynamic policy.
This standardized order may be modified by local configuration policy on Flow Specification filtering precedence, but if it does the BGP FlowSpecification Yang Model show indicate the current precedence.
Actions that change interact between levels of policy need to be defined in terms of policy actions in BGP Flow Specification. For example [I-D.litkowski-idr-flowspec-interfaceset] provides a definition of the following combination of filter rules between ACLs and BGP flow Specifications:
Pro - for version 2
The current version 1 of the Flow Specification does not have ordering of packet ECA policy rules, flow specification filters, or flow specification actions other than the default precedence. Current implementations of BGP flow specification are finding this lack of ordering to cause operational difficulties.
Con - for version 2
Version 2 must be coded. It can either be a BGP attribute with the policy rules (NLRI filters and actions) inside such as described in [I-D.li-idr-flowspec-rpd] or it can be a combination of a new BGP Flow Specification version 2 NLRI + Wide Community actions (with ordering).
(Additional comments will be added here)
The Flow Specification Yang models are expressing the same policy as the Filter-Based RIB Yang modules for I2RS and configuration. Aligning these three yang data models should improve the management of the different levels of filter-based forwarding (BGP Session ephemeral, I2RS reboot ephemeral, config filter-based forwarding).
This section compares BGP Flow Specification yang model in [I-D.wu-idr-flowspec-yang-cfg] and the I2RS FB-RIB data model is described in [I-D.hares-i2rs-fb-rib-data-model] which uses the packet reception ECA policy data model found in [I-D.hares-i2rs-pkt-eca-data-model]. A comparison of the policy structures is given in table 8, and the operation status model is given in table 9.
The packet reception ECA policy data model is also used to describe configured packet reception filter RIBs which (aka Policy Routing) described in (draft-hares-rtgwg-fb-rib-00.txt).
Table 8 - comparison of Yang Data models +-------------+----------------------+-----------------------+ | component | BGP Flow Spec | I2RS FB-RIB + | | | Yang | Packet-ECA Yang | +==============+=====================+=======================+ |Policy |flowspec-policy* |group* [group-name] | | +-name | [policy-name] | | | +-vrf |+-rw vrf-name | +-rw vrf-name | | +-AFI |+-rw address-family | +-rw address-famil | | +-rules |+-rw flowspec-rule* | +-rw group-rule-list | | || [rule-name] | | [rule-name] | | +-rule-name ||+-rw rule-name | |+-rw rule-name | | +-rule-order||+-rw traffic-filters| |+-rw rule-order | | ||+-rw traffic-actions| +-rw eca-rules | | | | | [order-id rule-name]| | | | | +-rw installer | | | | | +-rw eca-matches | | | | | +-rw eca-qos-actions| | | | | +-rw eca-fwd-actions| +--------------+---------------------+-----------------------+ figure 14 - Comparison of Yang modules (Config state)
Note:The Yang "traffic-filters" found are the same as eca-matches found in [I-D.wu-idr-flowspec-yang-cfg] are the same filters found in [I-D.hares-i2rs-pkt-eca-data-model]. The "traffic actions" found in [I-D.wu-idr-flowspec-yang-cfg] can be broken into modify actions and forwarding actions as [I-D.hares-i2rs-pkt-eca-data-model] does.
Table 9 - comparison of Yang operational state +------------+----------------------+-------------------------+ | component | BGP Flow Spec | I2RS FB-RIB | | | Yang | Packet-ECA Yang | +============+======================+=========================+ |opstate |flowspec-state |ietf-fb-ribs-oper-status | | +-rib |+-ro flowspec-rib |+-ro fb-rib-oper-status* | | | | | +-ro fb-rib-name | | +-groups | | | +-ro group-status | | +-rules | +-ro flowspec-entry*| +-ro rules_opstate | | [index] | [index] | [rule-order, rule-name]| | | | | | |statistics | | | | | +-rules |+-ro flowspec-stats* | +-ro rules_opstats | | | | [rule-order, rule-name]| | | +-ro vrf-name | | | | +-ro address-family | | | | +-ro flowspec-rule- | | | | | stats | | | | | | | | | |+-ro traffic-filters| | | | |+-ro traffic-action | | | | |+-ro classified-pkts| | +--ro pkts-match | | | | | | +--ro pkts-modified | | | |+-ro drop-pkts | | +--ro pkts-dropped | | | |+-ro drop-bytes | | +--ro bytes-dropped | | | | | +--ro pkts-forwarded | | | | | +--ro bytes-forwarded| +------------+----------------------+-------------------------+ figure 15 - Comparison of Yang Models (Operation State)
This section complies with [RFC7153]
TBD. There are a lot of assignments which will be filled in after the initial review of the technology.
The new BGP Validation described in section 4.1 with the ROA improves on [RFC5575] security by improving the validation of the originating AS having permissions to send Flow specifcation for a prefix. The validation of the path attributes and/or path requires the BGPSEC [I-D.ietf-sidr-bgpsec-protocol]. [I-D.eddy-idr-flowspec-exp] contains suggestions on how to implement this with flow specification, but at this time the authors consider the technology described in [I-D.eddy-idr-flowspec-exp] so this draft does not suggest mandating it. However, it encourages the develop of such work that pairs BGP Flow Specification with BGPSEC protocol. When this work matures, this specification or BGP Flow Specification version 2 should implement it.
| [I-D.eddy-idr-flowspec-exp] | Eddy, W., Dailey, J. and G. Clark, "Experimental BGP Flow Specification Extensions", Internet-Draft draft-eddy-idr-flowspec-exp-00, August 2015. |
| [I-D.eddy-idr-flowspec-packet-rate] | Eddy, W., Dailey, J. and G. Clark, "BGP Flow Specification Packet-Rate Action", Internet-Draft draft-eddy-idr-flowspec-packet-rate-00, November 2015. |
| [I-D.hao-idr-flowspec-nvo3] | Weiguo, H., Zhuang, S., Li, Z. and R. Gu, "Dissemination of Flow Specification Rules for NVO3", Internet-Draft draft-hao-idr-flowspec-nvo3-03, December 2015. |
| [I-D.hao-idr-flowspec-redirect-tunnel] | Weiguo, H., Li, Z. and L. Yong, "BGP Flow-Spec Redirect to Tunnel action", Internet-Draft draft-hao-idr-flowspec-redirect-tunnel-00, October 2015. |
| [I-D.hares-i2rs-fb-rib-data-model] | Hares, S., Kini, S., Dunbar, L., Ghanwani, A., Krishnan, R., Bogdanovic, D., Tantsura, J. and R. White, "Filter-Based RIB Data Model", Internet-Draft draft-hares-i2rs-fb-rib-data-model-01, January 2016. |
| [I-D.hares-i2rs-pkt-eca-data-model] | Hares, S., Wu, Q. and R. White, "Filter-Based Packet Forwarding ECA Policy", Internet-Draft draft-hares-i2rs-pkt-eca-data-model-00, January 2016. |
| [I-D.ietf-i2rs-architecture] | Atlas, A., Halpern, J., Hares, S., Ward, D. and T. Nadeau, "An Architecture for the Interface to the Routing System", Internet-Draft draft-ietf-i2rs-architecture-12, December 2015. |
| [I-D.ietf-i2rs-ephemeral-state] | Haas, J. and S. Hares, "I2RS Ephemeral State Requirements", Internet-Draft draft-ietf-i2rs-ephemeral-state-02, September 2015. |
| [I-D.ietf-idr-flow-spec-v6] | Raszuk, R., Pithawala, B., McPherson, D. and A. Andy, "Dissemination of Flow Specification Rules for IPv6", Internet-Draft draft-ietf-idr-flow-spec-v6-06, November 2014. |
| [I-D.ietf-idr-flowspec-l2vpn] | Weiguo, H., Litkowski, S. and S. Zhuang, "Dissemination of Flow Specification Rules for L2 VPN", Internet-Draft draft-ietf-idr-flowspec-l2vpn-03, November 2015. |
| [I-D.ietf-idr-wide-bgp-communities] | Raszuk, R., Haas, J., Lange, A., Amante, S., Decraene, B., Jakma, P. and R. Steenbergen, "Wide BGP Communities Attribute", Internet-Draft draft-ietf-idr-wide-bgp-communities-01, November 2015. |
| [I-D.ietf-netconf-restconf] | Bierman, A., Bjorklund, M. and K. Watsen, "RESTCONF Protocol", Internet-Draft draft-ietf-netconf-restconf-09, December 2015. |
| [I-D.ietf-netmod-acl-model] | Bogdanovic, D., Koushik, K., Huang, L. and D. Blair, "Network Access Control List (ACL) YANG Data Model", Internet-Draft draft-ietf-netmod-acl-model-06, December 2015. |
| [I-D.ietf-netmod-opstate-reqs] | Watsen, K. and T. Nadeau, "Terminology and Requirements for Enhanced Handling of Operational State", Internet-Draft draft-ietf-netmod-opstate-reqs-04, January 2016. |
| [I-D.ietf-netmod-routing-cfg] | Lhotka, L. and A. Lindem, "A YANG Data Model for Routing Management", Internet-Draft draft-ietf-netmod-routing-cfg-20, October 2015. |
| [I-D.ietf-sidr-bgpsec-protocol] | Lepinski, M., "BGPsec Protocol Specification", Internet-Draft draft-ietf-sidr-bgpsec-protocol-14, December 2015. |
| [I-D.kini-i2rs-fb-rib-info-model] | Kini, S., Hares, S., Dunbar, L., Ghanwani, A., Krishnan, R., Bogdanovic, D., Tantsura, J. and R. White, "Filter-Based RIB Information Model", Internet-Draft draft-kini-i2rs-fb-rib-info-model-02, October 2015. |
| [I-D.li-idr-flowspec-rpd] | Li, Z., Ou, L., Luo, Y., Lu, S., Zhuang, S. and N. Wu, "BGP FlowSpec Extensions for Routing Policy Distribution (RPD)", Internet-Draft draft-li-idr-flowspec-rpd-01, October 2015. |
| [I-D.liang-idr-bgp-flowspec-label] | You, J., Raszuk, R. and d. danma@cisco.com, "Carrying Label Information for BGP FlowSpec", Internet-Draft draft-liang-idr-bgp-flowspec-label-01, September 2015. |
| [I-D.liang-idr-bgp-flowspec-time] | You, J. and S. Zhuang, "BGP FlowSpec with Time Constraints", Internet-Draft draft-liang-idr-bgp-flowspec-time-00, October 2015. |
| [I-D.litkowski-idr-flowspec-interfaceset] | Litkowski, S., Simpson, A., Patel, K. and J. Haas, "Applying BGP flowspec rules on a specific interface set", Internet-Draft draft-litkowski-idr-flowspec-interfaceset-03, December 2015. |
| [I-D.rosen-idr-tunnel-encaps] | Rosen, E., Patel, K. and G. Velde, "Using the BGP Tunnel Encapsulation Attribute without the BGP Encapsulation SAFI", Internet-Draft draft-rosen-idr-tunnel-encaps-03, August 2015. |
| [I-D.vandevelde-idr-flowspec-path-redirect] | Velde, G., Henderickx, W. and K. Patel, "Flowspec Indirection-id Redirect", Internet-Draft draft-vandevelde-idr-flowspec-path-redirect-01, January 2016. |
| [I-D.wu-idr-flowspec-yang-cfg] | Wu, N., Zhuang, S. and A. Choudhary, "A YANG Data Model for Flow Specification", Internet-Draft draft-wu-idr-flowspec-yang-cfg-02, October 2015. |
| [RFC4303] | Kent, S., "IP Encapsulating Security Payload (ESP)", RFC 4303, DOI 10.17487/RFC4303, December 2005. |