idnits 2.17.1 draft-herbert-fast-02.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** The document seems to lack a both a reference to RFC 2119 and the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords. RFC 2119 keyword, line 428: '... o Tickets MUST be stateless with...' RFC 2119 keyword, line 429: '...termediate nodes MUST NOT be required ...' RFC 2119 keyword, line 432: '... o Tickets MUST work in a multi-h...' RFC 2119 keyword, line 434: '... that issued a ticket, tickets MUST be...' RFC 2119 keyword, line 438: '... o Tickets MUST work with any tra...' (28 more instances...) Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (July 24, 2018) is 2096 days in the past. Is this intentional? Checking references for intended status: Full Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) ** Obsolete normative reference: RFC 5226 (Obsoleted by RFC 8126) == Outdated reference: A later version (-03) exists of draft-herbert-6man-icmp-limits-00 Summary: 2 errors (**), 0 flaws (~~), 2 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 INTERNET-DRAFT T. Herbert 3 Intended Status: Standard Quantonium 4 Expires: January 2019 6 July 24, 2018 8 Firewall and Service Tickets 9 draft-herbert-fast-02 11 Abstract 13 This document describes the Firewalls and Service Tickets protocol. A 14 ticket is data that accompanies a packet and indicates a granted 15 right to traverse a network or a request for network service to be 16 applied. Applications request tickets from a local agent in the 17 network and attach issued tickets to packets. Firewall tickets are 18 issued to grant packets the right to traverse a network; service 19 tickets indicate the desired service to be applied to a packets. A 20 single ticket may provide both firewall and service ticket 21 information. Tickets are sent in IPv6 Hop-by-Hop options. 23 Status of this Memo 25 This Internet-Draft is submitted to IETF in full conformance with the 26 provisions of BCP 78 and BCP 79. 28 Internet-Drafts are working documents of the Internet Engineering 29 Task Force (IETF), its areas, and its working groups. Note that 30 other groups may also distribute working documents as 31 Internet-Drafts. 33 Internet-Drafts are draft documents valid for a maximum of six months 34 and may be updated, replaced, or obsoleted by other documents at any 35 time. It is inappropriate to use Internet-Drafts as reference 36 material or to cite them other than as "work in progress." 38 The list of current Internet-Drafts can be accessed at 39 http://www.ietf.org/1id-abstracts.html 41 The list of Internet-Draft Shadow Directories can be accessed at 42 http://www.ietf.org/shadow.html 44 Copyright and License Notice 45 Copyright (c) 2018 IETF Trust and the persons identified as the 46 document authors. All rights reserved. 48 This document is subject to BCP 78 and the IETF Trust's Legal 49 Provisions Relating to IETF Documents 50 (http://trustee.ietf.org/license-info) in effect on the date of 51 publication of this document. Please review these documents 52 carefully, as they describe your rights and restrictions with respect 53 to this document. Code Components extracted from this document must 54 include Simplified BSD License text as described in Section 4.e of 55 the Trust Legal Provisions and are provided without warranty as 56 described in the Simplified BSD License. 58 Table of Contents 60 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 61 2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . 4 62 2.1 Current mechanisms . . . . . . . . . . . . . . . . . . . . . 4 63 2.1.1 Stateful firewalls and proxies . . . . . . . . . . . . . 4 64 2.1.2 QoS signaling . . . . . . . . . . . . . . . . . . . . . 5 65 2.1.3 Deep packet inspection . . . . . . . . . . . . . . . . . 5 66 2.2 Proposals for applications to signal the network . . . . . . 5 67 2.2.1 SPUD/PLUS . . . . . . . . . . . . . . . . . . . . . . . 5 68 2.2.2 Path aware networking . . . . . . . . . . . . . . . . . 7 69 2.3 Emerging use cases . . . . . . . . . . . . . . . . . . . . . 7 70 3 Architecture . . . . . . . . . . . . . . . . . . . . . . . . . 8 71 3.1 Example packet flow . . . . . . . . . . . . . . . . . . . . 9 72 3.2 Requirements . . . . . . . . . . . . . . . . . . . . . . . . 10 73 4 Packet format . . . . . . . . . . . . . . . . . . . . . . . . . 11 74 4.1 Option format . . . . . . . . . . . . . . . . . . . . . . . 11 75 4.2 Option types . . . . . . . . . . . . . . . . . . . . . . . . 12 76 4.3 Ticket format . . . . . . . . . . . . . . . . . . . . . . . 12 77 5 Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 78 5.1 Ticket types and ordering . . . . . . . . . . . . . . . . . 13 79 5.2 Origin application processing . . . . . . . . . . . . . . . 14 80 5.2.1 Ticket requests . . . . . . . . . . . . . . . . . . . . 14 81 5.2.2 Ticket identification . . . . . . . . . . . . . . . . . 14 82 5.2.3 Ticket use . . . . . . . . . . . . . . . . . . . . . . . 14 83 5.2.4 Ticket scope . . . . . . . . . . . . . . . . . . . . . . 15 84 5.3 Origin network processing . . . . . . . . . . . . . . . . . 15 85 5.4 Peer host processing . . . . . . . . . . . . . . . . . . . . 16 86 5.5 Processing reflected tickets . . . . . . . . . . . . . . . . 17 87 5.5.1 Network processing . . . . . . . . . . . . . . . . . . . 17 88 5.5.2 Host processing . . . . . . . . . . . . . . . . . . . . 17 89 5.6 Handling dropped extension headers . . . . . . . . . . . . . 17 90 5.6.1 Mitigation for dropped extension headers . . . . . . . . 17 91 5.6.2 Fallback for dropped extension headers . . . . . . . . . 18 93 6 Implementation considerations . . . . . . . . . . . . . . . . . 19 94 6.1 Origin applications . . . . . . . . . . . . . . . . . . . . 19 95 6.2 Ticket reflection . . . . . . . . . . . . . . . . . . . . . 19 96 7 Security Considerations . . . . . . . . . . . . . . . . . . . . 19 97 8 IANA Considerations . . . . . . . . . . . . . . . . . . . . . . 20 98 9 References . . . . . . . . . . . . . . . . . . . . . . . . . . 20 99 9.1 Normative References . . . . . . . . . . . . . . . . . . . 20 100 9.2 Informative References . . . . . . . . . . . . . . . . . . 21 101 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 21 103 1 Introduction 105 Firewall and Service Tickets (FAST) is a technique to allow an 106 application to signal to the network requests for admission and 107 services for packets. A ticket is data that is attached to a packet 108 by the source node, and it is then inspected and validated by certain 109 intermediate nodes in a network. Tickets express a grant or right for 110 packets to traverse a network or have services applied to them. 112 An application requests tickets for admission or services from a 113 ticket agent in their local network. The agent issues tickets to the 114 application which in turn attaches these to its packets. In the 115 forwarding path, intermediate network nodes interpret tickets and 116 apply requested services on packets. 118 Tickets are validated for authenticity by the network and contain an 119 expiration time so that they cannot be easily forged. Tickets do not 120 have a global interpretation, they can only be interpreted within the 121 network that issues them. In order to apply services to inbound 122 packets for a communication, remote peers reflect received tickets in 123 packets they send without interpreting them. Tickets are stateless 124 within the network, however they can be used to attain per flow 125 semantics. Firewall and service tickets are are non-transferable and 126 revocable. 128 Tickets are coded in IPv6 Hop-by-Hop options. 130 2 Motivation 132 This section presents the motivation for Firewall and Service 133 Tickets. 135 2.1 Current mechanisms 137 Current solutions for controlling admission to the network and 138 requesting services are mostly ad hoc and architecturally limiting. 140 2.1.1 Stateful firewalls and proxies 142 Stateful firewalls and proxies are the predominantly deployed 143 techniques to control access to a network on a per flow basis. While 144 they provide some benefits of security, they break the end-to-end 145 model and have otherwise restricted the Internet in several ways: 147 o They require parsing over transport layer headers in the fast 148 path of forwarding. 150 o They are limited to work only with a handful of protocols and 151 protocol features thereby ossifying protocols. 153 o They break the ability to use mulithoming and multipath. All 154 packets for a flow must traverse a specific network device in 155 both directions of a communication. 157 o They can break end to end security. NAT for instance breaks the 158 TCP authentication option. 160 o They have created single points of failure and become network 161 bottlenecks. 163 2.1.2 QoS signaling 165 In the current Internet, there is little coordination between hosts 166 and the network to provide services based on characteristics of the 167 application. Differentiated services provides an IP layer means to 168 classify and manage traffic, however it is lacking in richness of 169 expression and lacks a ubiquitous interface that allows applications 170 to request service with any granularity. Without additional state, 171 there is no means for the network infrastructure to validate that a 172 third party application has requested QoS that adheres to network 173 policies. 175 2.1.3 Deep packet inspection 177 Some network devices perform Deep Packet Inspection (DPI) into the 178 application data to determine whether to admit packets or what 179 services to apply. For instance, HTTP is commonly parsed to determine 180 URL, content type, and other application level information the 181 network is interested in. DPI can only be effective with the 182 application layer protocols that a device is programmed to parse. 183 More importantly, application level DPI is being effectively 184 obsoleted in the network due the pervasive use of TLS. TLS 185 interception and SSL inspection, whereby an intermediate node 186 implements a proxy that decrypts a TLS session and re-encrypts, is 187 considered a security vulnerability [TLSCERT]. 189 2.2 Proposals for applications to signal the network 191 This section surveys some proposals to address the need for 192 applications to signal the network. 194 2.2.1 SPUD/PLUS 196 SPUD (Session Protocol Underneath Datagrams) [SPUD] and its successor 197 PLUS (Path Layer UDP Substrate) [PLUS] proposed a UDP based protocol 198 to allow applications to signal a rich set of characteristics and 199 service requirements to the network. 201 SPUD had a number of drawbacks: 203 o SPUD is based on a specific protocol used over UDP. This requires 204 applications to change to use a new protocol. In particular SPUD 205 is incompatible with TCP which is the predominant transport 206 protocol on the Internet. 208 o SPUD requires that intermediate nodes parse and process UDP 209 payloads. Since UDP port numbers do not have global meaning 210 [RFC7605] there is the possibility of misinterpretation and of 211 silent data corruption if intermediate nodes modify UDP payloads. 212 SPUD attempts to mitigate this issue with the use of magic 213 numbers, however that can only ever be probabilistically correct. 215 o SPUD included stateful flow tracking in the network. This 216 problematic because: 218 o Not all communications have well defined connection semantics. 219 For instance, a unidirectional data stream has no connection 220 semantics at all. 222 o Stateful network devices breaks multi-homing and multi-path; 223 they assume that all packets of a flow in both directions are 224 seen by the node doing tracking flow state. Stateful 225 firewalls, for instance, require all packets for a flow to 226 always go through the same device in both directions. This 227 disallows flexibility and optimized traffic flow that a multi- 228 homed network affords. 230 o Maintaining per flow state in the network is an obvious 231 scaling problem. 233 o Keepalives to maintain a network state in a device, such as 234 those sent to prevent a NAT state from being evicted, carry no 235 useful information to the end user and in large numbers can 236 become a source of congestion. 238 o The meta data information in SPUD would have global definition. 239 This problematic because: 241 o Application specific information could be leaked to unknown 242 and untrusted parties. 244 o Establishing a specification on what data should be conveyed 245 in SPUD will be difficult. Different service providers may 246 want different pieces of information, applications may also 247 have different ideas about what information is safe to make 248 visible. 250 2.2.2 Path aware networking 252 Path aware networking (PAN) [PAN] is an IRTF effort to allow 253 applications to select paths through the Internet for their traffic. 254 The idea is that in choosing different paths an application can 255 select from the different characteristics that are associated with 256 different paths. Path aware networking requires a means to express 257 the various paths and their associated characteristics to 258 applications. In the data path, a method to express desired path for 259 a packet is needed-- this presumably could be by a mechanism such as 260 segment routing. 262 PAN and FAST have similar characteristics, particularly with respect 263 to the need for applications and networks to communicate about 264 network path characteristics. However, where PAN presumably endeavors 265 to allow path selection by an application, FAST allows applications 266 to select their desired path characteristics and it is up to the 267 network to select the actual path. This distinction is important to 268 maximize flexibility, especially in situations where providing any 269 detailed path information to untrusted end device is a security risk 270 (which would be the typical case in a provider network or open 271 Internet). 273 2.3 Emerging use cases 275 In a typical client/server model of serving content, end host clients 276 communicate with servers on the Internet. Clients are typically user 277 devices that are connected to the Internet through a provider 278 network. In the case of mobile devices, such as smart phones, the 279 devices are connected to the Internet through a mobile provider 280 network. Content providers (web servers and content caches) tend to 281 be more directly connected to the Internet, the largest of which can 282 connect at exchange points. 284 Provider networks can be architected to provide different services 285 and levels of services to their users based on characteristics of 286 applications. For example, a mobile carrier network can provide 287 different latency and throughput guarantees for different types of 288 content. A network may offer different services for optimizing video: 289 streaming an HD movie might need high throughput but not particularly 290 low latency; a live video chat might have lower throughput demands 291 but have stringent low latency requirements. 293 The emerging 3GPP standard for 5G defines a set of mechanisms to 294 provide a rich array of services for users. These mechanisms employ 295 Network Function Virtulization (NFV), Service Function Chaining 296 (SFC), and network slices that divide physical network resources into 297 different virtualized slices to provide different services. To make 298 use of these mechanisms, the applications running in UEs (User 299 Equipment) will need to indicate desired services of the RAN (Radio 300 Access Network). For instance, a video chat application may request 301 bounded latency that is implemented by the network as a network 302 slice; so packets sent by the application should be mapped to that 303 network slice. 305 Note that an application service applies to both packets sent by an 306 end node and those sent from a peer towards the end node. For the 307 latter case, the network needs to be able to map packets sent from 308 hosts on the Internet to the services requested by the receiving 309 application. 311 3 Architecture 313 The figure below illustrates an example network path between two 314 hosts on the Internet. Each host connects to the Internet via a 315 provider network, and provider networks are connected in the Internet 316 by transit networks. 317 _____ 318 __________ ( ) __________ 319 +--------+ ( ) ( ) ( ) +--------+ 320 | User 1 +---( Provider A )--( Transit )--( Provider B )---+ User 2 | 321 +--------+ (__________) ( ) (__________) +--------+ 322 (_____) 324 Figure 1 326 Within each provider network, services may be provided on behalf of 327 the users of the network. In the figure above, Provider 1 may provide 328 services and service agreements for users in its network including 329 User 1; and likewise Provider B can provide services to users in its 330 network including User 2. Transit networks service all users and 331 don't typically provide user specific services or service 332 differentiation. 334 Services provided by different provider networks may be very 335 different and dependent on the implementation of the network as well 336 as the policies of the provider. 338 Based on this model, services and service differentiation can be 339 considered local to each network provider. FAST is a mechanism 340 whereby each user and application can request from its local provider 341 the services to be applied to its traffic. A request for service is 342 made to a FAST "ticket agent". The contents of the request describe 343 the services that application desires. The ticket agent responds with 344 a "ticket" that the application sets in its packets. When a packet is 345 sent by the application with a ticket attached, the ticket is 346 interpreted in the provider network to allow the packet to traverse 347 the network and to map the packet to the appropriate services. The 348 ticket is only relevant to the provider network that issued the 349 ticket, to the application itself and nodes outside of the provider 350 network the ticket is an uinterpretable opaque object. 352 To facilitate network traversal and service mapping in the reverse 353 direction for a flow, that is packets sent from a peer host, peer 354 hosts reflect tickets without modification or interpretation. This is 355 done by saving the ticket received in packets of a flow and attaching 356 that as a reflected ticket to packets being sent on the flow. 358 The use of tickets may be bilateral for a flow so that each peer 359 requests service from its local network. Therefore packets may 360 contain two types of tickets: one that is set by the sending host to 361 signal its local provider network, and the other is the reflected 362 ticket that is a signal to the provider network of the peer endpoint. 364 Tickets are scoped values, they only have meaning in the network in 365 which they were issued. The format, meaning, and interpretation of 366 tickets is network specific. By mutual agreement, two networks may 367 share the policy and interpretations of tickets. For instance, there 368 could be an agreement between two provider networks to interpret each 369 others tickets or to use a common format. 371 3.1 Example packet flow 373 Referencing the diagram in figure 1, consider that User 1 is 374 establishing a video chat with User 2 and wishes to have low latency 375 service for video applied by its local network (Provider 1). The flow 376 of events may be: 378 1. User 1 makes a ticket request to a ticket agent of Provider A 379 that describes the video application and may include detailed 380 characteristics such as resolution, frame rate, latency, etc. 382 2. The ticket agent issues a ticket to User 1 that indicates that 383 packets of the flow have a right to traverse the network and 384 the services to be applied to the packets of the flow. 386 3. The video chat application sends packets with the ticket 387 attached for the video chat. 389 4. The first hop node in Provider A's network interprets the 390 ticket in packets and applies the appropriate services (e.g. 392 sets diffserv, forwards on a network slice, encapsulates in 393 MPLS, etc.). 395 5. Packets traverse Provider A's network with the appropriate 396 services being applied. 398 6. Packets traverse transit networks and Provider B network, the 399 attached tickets are ignored. 401 7. Packets are received at User 2. Attached tickets are saved in 402 the context of the flow for the video chat. 404 8. User 2's video chat application sends packets to User 1. The 405 last ticket previously received from User 1 is now reflected in 406 these packets. 408 9. Packets traverse Provider B network and transit networks, the 409 reflected ticket is ignored. 411 10. An ingress node in Provider A's network interprets the 412 reflected ticket and applies appropriate services to the 413 packets for traversing the local network. 415 11. Packets are forwarded within Provider's A network with the 416 appropriate services applied. No other intermediate nodes 417 should need to interpret the tickets. 419 12. Packets are received at the host for User 1. The reflected 420 ticket is validated by comparing the received ticket with that 421 being sent for the flow. It the ticket is determined valid then 422 the packet is accepted. 424 3.2 Requirements 426 The requirements for Firewall and Service Tickets are: 428 o Tickets MUST be stateless within the network. In particular 429 intermediate nodes MUST NOT be required to create and maintain 430 state for transport layer connections. 432 o Tickets MUST work in a multi-homed and multi-path environments. 434 o Outside of the network that issued a ticket, tickets MUST be 435 opaque and obfuscated so that no application specific 436 information is derivable. 438 o Tickets MUST work with any transport protocol as well as in the 439 presence of any IP protocol feature (e.g. other extension 440 headers are present). 442 o Tickets SHOULD minimize the changes to an application. Their use 443 should be an "add-on" to the existing communications of an 444 application. 446 o Tickets MUST deter spoofing and other misuse that might result 447 in illegitimate use of network services or denial of service 448 attack. 450 o Tickets MUST be contained in the IP layer protocol. In 451 particular, tickets MUST NOT require parsing transport layer 452 headers. 454 o Tickets MUST allow services to be applied in the return path of 455 a communication. In a client/server application it is often the 456 packets in the reverse path that require the most service (for 457 instance if a video is being streamed to a client). 459 o A fallback MUST be present to handle the case that extension 460 headers are dropped within the network or a peer node does not 461 reflect tickets. A fallback allows functional communications but 462 provides it in a degrade mode of service. 464 4 Packet format 466 A ticket is sent in a Hop-by-Hop option. 468 4.1 Option format 470 The format of an Hop-by-Hop option containing a ticket is: 472 1 2 3 473 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 1 474 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 475 | Option Type | Opt Data Len | Type | | 476 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + 477 | | 478 ~ Ticket ~ 479 | | 480 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 482 Fields: 484 o Option type: Type of Hop-by-Hop option. This document proposes 485 two possible values for ticket: an unmodifiable and a modifiable 486 variant. 488 o Opt Data Len: Length of the option data field. This is two plus 489 the length of the ticket field 491 o Type: Indicates the type and reflection property of the ticket. 492 Possible values are: 494 o 0x0: Ticket from origin, don't reflect at receiver 496 o 0x1: Ticket from origin, reflect at receiver 498 o 0x2: Reflected ticket 500 o 0x3-0xf: Reserved 502 4.2 Option types 504 The are two option numbers requested for the ticket option: 0x0F and 505 0x2F. The latter allows modification. This would be used in 506 situations where the network nodes needed to modify the option with 507 new information. 509 4.3 Ticket format 511 A ticket encodes service parameters that describe the desired 512 services as well as additional fields that would be used to provide 513 privacy and integrity. 515 The format of a ticket is defined by the network in which the ticket 516 is issued. A ticket should be obfuscated or encrypted for privacy so 517 that only the local network can interpret it. It should be 518 uniterpretable to any nodes outside the network and to the 519 application or host that is granted a ticket. It should be resistant 520 to spoofing so that an attacker cannot illegitimately get service by 521 applying a ticket seen on other flows. 523 It is recommended that tickets are encrypted and each ticket has an 524 expiration time. For instance, a ticket may be created by encrypting 525 the ticket data with an expiration time and using the source address, 526 destination address, and a shared key as the key for encryption. 528 For example, a ticket with an expiration time may have the format: 530 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 1 531 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 532 | Expiration time | 533 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 534 | | 535 ~ Service parameters ~ 536 | | 537 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 539 Where the expiration time is in a format understood by the local 540 network nodes which maintain synchronized time. The Service 541 parameters are relevant to network nodes and describe the services to 542 be applied. The service parameters could simply be a set of flags for 543 services, an index to a service profile known by the network nodes, 544 or possibly have more elaborate structure that could indicate 545 numerical values for characteristics that have a range. The service 546 parameters could also include a type field to allow a network to 547 define different representations of service parameters. 549 A simple ticket containing a service protocol index might be: 551 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 1 552 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 553 | Expiration time | 554 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 555 | Type | Service Profile Index | 556 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 558 Where Type indicates the type of ticket and in this case indicates it 559 is a service profile index. Service Profile Index could be an index 560 into a table that describes the services to be applied. 562 5 Operation 564 5.1 Ticket types and ordering 566 There are three types of tickets that may be contained in a packet: 568 o origin tickets that are not reflected 570 o origin tickets to be reflected 572 o reflected tickets 574 Origin tickets are those set by an application that was issued a 575 ticket and have an additional property indicating whether they are to 576 be reflected by a peer host. Reflected tickets are those that have 577 been received and reflected by a peer host. 579 A sender SHOULD set at most one of each type of ticket in a packet. 580 If different types of ticket are sent within a single packet they 581 SHOULD have the following ordering: 583 1. origin tickets that are not reflected 584 2. origin tickets to be reflected 586 3. reflected tickets 588 If a packet contains more than one ticket option with the same ticket 589 type, only the first option appearing in the list for the ticket type 590 is processed. Additional options for the same ticket type are parsed 591 but not processed. 593 5.2 Origin application processing 595 An origin application requests tickets, sets them in packets, and 596 validates reflected tickets. 598 5.2.1 Ticket requests 600 An application that wishes to use network services first requests 601 tickets from a ticket agent. The application request could be in the 602 form of an XML structure with a canonical elements (the definition is 603 outside the scope of this document). The application makes a request 604 to the ticket agent for the local network. This could be done via a 605 web service using REST APIs. Internally in the host, the ticket agent 606 might be accessed through a library that interfaces to a ticket 607 daemon that in turn arbitrates requests between the applications and 608 a ticket agent in the network. 610 An issued ticket is opaque to the application and the application 611 should not attempt to interpret it or take any other action other 612 than attaching the ticket to its packets. 614 A ticket agent MAY provide both a origin ticket not to be reflected 615 and one that is to be reflected. The intent is that different tickets 616 can be used between the outbound and inbound paths for the flow. In 617 the case that two tickets are provided, the origin ticket not to be 618 reflected MUST appear first in the options list. 620 5.2.2 Ticket identification 622 Tickets are valid for a specific IP source and destination address 623 for which they were issued. Transport layer ports and other transport 624 layer information are not included ticket identification, however an 625 application can request tickets and validate reflected tickets on a 626 per flow basis. Issued tickets are stored in the flow context and the 627 saved information is used to validate reflected tickets. 629 5.2.3 Ticket use 631 When the ticket agent issues an returns a ticket, the application 632 sets the ticket as a Hop-by-Hop option. This is typically done by 633 setting a socket option on a socket (in the case of TCP) or by 634 indicating the option in the ancillary data when sending on a 635 unconnected socket (in the case of UDP). The application SHOULD 636 continue to use the same ticket for the flow until it is updated with 637 a new ticket. 639 The ticket agent SHOULD return an expiration time with the ticket. An 640 application can use the ticket until the expiration time, at which 641 point it can request a new ticket to continue communications. In 642 order to make the ticket transition process seamless an application 643 MAY request a new ticket before the old one expires. 645 A host SHOULD set in a packet at most one origin ticket not to be 646 reflected and one origin ticket to be reflected. 648 5.2.4 Ticket scope 650 Tickets are valid for a specific IP source and destination address 651 for which they were issued. Transport layer ports and other transport 652 layer information are not included ticket identification, however an 653 application can request tickets and validate reflected tickets on a 654 per flow basis. Issued tickets are stored in the flow context and the 655 saved information is used to validate reflected tickets. 657 A network MAY delegate creation of tickets to hosts in a limited 658 fashion. This would entail the network ticket agent issuing a master 659 ticket to a host ticket agent which in turn can use the master ticket 660 to create a limited number of tickets. The details of ticket agent 661 delegation are outside the scope of this document. 663 5.3 Origin network processing 665 When a packet with a ticket enters a network, a network node can 666 determine if the ticket originated in its network and must be 667 processed. This is done by considering the origin of the ticket and 668 the source or destination IP address. For an origin ticket (i.e. 669 ticket is not reflected), the source address is considered. If the 670 source address is local to the network then the ticket can be 671 interpreted. For a reflected ticket, the destination address is 672 considered. If the destination address is local to the network then 673 the ticket can be interpreted. 675 If a ticket origin is determined to be the local network then the 676 ticket is processed. The ticket is decrypted if necessary and the 677 expiration time is checked. If the ticket is verified to be authentic 678 and valid then the packet is mapped to be processed by the requested 679 services. For instance, in a 5G network the packet may be forwarded 680 on a network slice for the characteristics the application has 681 requested (real-time video for instance). 683 If an origin ticket cannot be verified, for instance the ticket 684 cannot be authenticated, then the ticket SHOULD be ignored and the 685 packet processed as though no ticket were present. 687 Note that there are logically only two ingress points into the 688 network at which a provider needs to process tickets: when a local 689 user sends a packet into the provider network with an origin ticket, 690 and when a packet from an external network enters the provider's 691 network with a reflected ticket. Any ticket should be processed at 692 most once within a network. Once a ticket is processed and mapped to 693 the network's service mechanisms it should not need further 694 examination. 696 If there is more than one origin ticket present, then the first one 697 encountered is processed and any additional origin tickets SHOULD be 698 ignored by a network node. Note that this will be the case if a 699 ticket agent issued both a origin ticket not to be reflected and one 700 to be reflected; the ticket not to be reflected should appear first 701 in the packet and thus would be the one processed by the node. 703 If there is more than one reflected ticket present, then the first 704 one encountered is processed and any additional reflected tickets 705 SHOULD be ignored. 707 5.4 Peer host processing 709 When a host receives a packet with a ticket whose type is "from 710 origin and needs to be reflected", it SHOULD save the ticket in its 711 flow context and reflect it on subsequent packets. When the 712 application reflects the option, it copies the whole option and only 713 modifies the type to indicate a "reflected ticket". The application 714 SHOULD continue to reflect the ticket until a different one is 715 received from the origin or a packet without a service ticket option 716 is received on the flow. Note that the latest ticket that is received 717 is the one to be reflected, if packets have been received out of 718 order for a flow it is possible that the reflected ticket is from an 719 earlier packet in a flow. 721 If there is more than one origin ticket to be reflected present, then 722 the first one encountered is processed and any additional origin 723 tickets to be reflected SHOULD be ignored. 725 A peer host MUST ignore received origin tickets that are not to be 726 reflected. 728 5.5 Processing reflected tickets 730 5.5.1 Network processing 732 When a packet with a reflected ticket enters the origin network of 733 the ticket, the ticket MUST be processed. The ticket is validated. 734 Validation entails decoding or decrypting the ticket and checking the 735 expiration time. If the ticket is valid and has not expired time then 736 the packet is verified for forwarding. 738 A network MAY accept expired reflected tickets for some configurable 739 period after the expiration time. Rate limiting SHOULD be applied to 740 packets with expired reflected tickets. Accepting expired tickets is 741 useful in the case that a connection goes idle and after sometime the 742 remote peer starts to send. The ticket it reflects may be expired and 743 presumably the receiving host will quickly respond with a new ticket. 745 5.5.2 Host processing 747 Upon receiving a packet with a reflected ticket an end host MUST 748 validate the ticket before accepting the packet. This verification is 749 done by comparing the received ticket to that which is set to be sent 750 on the corresponding flow. If the tickets do not match then the 751 packet is dropped and the event SHOULD be logged. 753 A host SHOULD retain and validate expired tickets that are reflected 754 to allow a peer time to receive and reflect an updated ticket. 756 5.6 Handling dropped extension headers 758 The downside of using IPv6 extension headers on the Internet is that 759 they are currently not completely reliable. Some intermediate nodes 760 will drop extension headers with rates described in [RFC7872]. 762 5.6.1 Mitigation for dropped extension headers 764 There are some mitigating factors for this problem: 766 o A provider network that implements tickets would need to ensure 767 that extension headers are at least usable within their network. 769 o Transit networks are less likely to arbitrarily drop packets 770 with extension headers. 772 o Many content providers, especially the larger ones, may be 773 directly connected to the Internet. For example, front end web 774 servers may be co-located as exchange points. 776 o The requirement that nodes must process Hop-by-hop options has 777 been relaxed in [RFC8200]. It is permissible for intermediate 778 nodes to ignore them. 780 o Increased deployment of IPv6 and viable use cases of extension 781 headers, such as described here, may motivate vendors to fix 782 issues with extension headers. 784 5.6.2 Fallback for dropped extension headers 786 Since the possibility that extension headers are dropped cannot be 787 completely eliminated, a fallback is included for use with tickets. 789 When an application connects to a new destination for which it has no 790 history about the viability of extension headers, it can perform a 791 type of Happy Eyeballs probing. The concept is for a host to send a 792 number of packets with and without tickets. The application can 793 observe whether packets with tickets are being dropped or not being 794 reflected. 796 There are a few possible outcomes of this process: 798 o A packet with a ticket is dropped and an ICMP for extension 799 headers [ICMPEH] processing limits is received. This is a strong 800 signal that extension headers are not viable to the destination 801 and should not be used for the flow. 803 o A packet with a ticket is dropped and no ICMP error is received. 804 This is a signal that extension headers may not be usable. If 805 such drops are observed for all or a significant fraction of 806 packets and there are no drops for packets that were sent 807 without tickets, then extension headers should be considered not 808 viable for the flow. 810 o Packets with tickets are not being dropped, however tickets are 811 not being reflected. This is a signal that the peer application 812 does not support reflection. Tickets may be sent, however they 813 are only useful in the outbound path. 815 o Packets with tickets are not being dropped and tickets are 816 properly being reflected. Tickets are useful in both directions. 818 If extension headers are found to not be viable or tickets are not 819 being properly reflected, a possible fallback is to not use tickets. 820 In this case, communications might remain functional, however they 821 would be operate in a degraded mode of service. The network may 822 fallback to creating per flow state in the network; the ticket that 823 an application sent with packets during probing could be used to 824 instantiate the service characteristics maintained in a flow state. 826 6 Implementation considerations 828 6.1 Origin applications 830 Existing client applications can be modified to request tickets and 831 set them in packets. The kernel may need some small changes or 832 configuration to enable an application to specify the option for its 833 packets. 835 The interface to the ticket agent would likely be via a library API. 837 Using the BSD sockets interface, for a connected socket (TCP, SCTP, 838 or connected UDP socket) a Hop-by-Hop option can be set on the socket 839 via the setsockopt system call. For an unconnected socket (UDP) the 840 ticket option can be set as ancillary data in the sendmsg system 841 call. 843 Happy Eyeballs for extension headers, described in section 5.6.2, 844 could be implemented in the kernel for a connection oriented 845 transport protocol such a TCP. For connectionless protocols, probing 846 could be handled by an application library. 848 6.2 Ticket reflection 850 To perform ticket reflection, servers must be updated. In the case of 851 a connected socket (TCP, SCTP, or a connected UDP socket) this can be 852 done as relatively minor change to the kernel networking stack which 853 would be transparent to applications. For unconnected UDP, an 854 application could receive the ticket as part of the ancillary data in 855 recvmsg system call, and then send the reflected ticket in a reply 856 using ancillary data in sendmsg. 858 7 Security Considerations 860 There are two main security considerations: 862 o Leakage of content specific information to untrusted third 863 parties must be avoided. 865 o Tickets cannot be forged or illegitimately used. 867 Tickets may be visible to the Internet including untrusted and 868 unknown networks in the path of sent packets. Therefore, tickets 869 should be encrypted or obfuscated by the origin network. 871 Tickets need to have an expiration time, must be resistant to 872 forgery, and must be nontransferable. A ticket should be valid for 873 the specific source and destination addresses that it was issued for. 874 Tickets are revocable by implemented a black-list contained revoked 875 tickets. 877 8 IANA Considerations 879 IANA is requested to assigned the following Hop-By-Hop options: 881 +-----------+---------------+-------------+---------------+ 882 | Hex Value | Binary value | Description | Reference | 883 | | act chg rest | | | 884 +-----------+---------------+-------------+---------------+ 885 | 0x0F | 00 0 01111 | Firewall | This document | 886 | | | and Service | | 887 | | | Ticket | | 888 +-----------+---------------+-------------+---------------+ 889 | 0x2F | 00 1 01111 | Modifiable | This document | 890 | | | Firewall | | 891 | | | and Service | | 892 | | | Ticket | | 893 +-----------+---------------+-------------+---------------+ 895 IANA is requested to set up a registry for the Ticket types. These 896 types are 4 bit values. New values for control types 0x3-0xf are 897 assigned via Standards Action [RFC5226]. 899 +----------------+--------------------+---------------+ 900 | Ticket type | Description | Reference | 901 +----------------+--------------------+---------------+ 902 | 0x0 | Ticket from origin | This document | 903 | | and don't reflect | | 904 +----------------+--------------------+---------------+ 905 | 0x1 | Ticket from origin | This document | 906 | | and reflect | | 907 +----------------+--------------------+---------------+ 908 | 0x2 | Reflected ticket | This document | 909 +----------------+--------------------+---------------+ 911 9 References 913 9.1 Normative References 915 [RFC8200] Deering, S. and R. Hinden, "Internet Protocol, Version 6 916 (IPv6) Specification", STD 86, RFC 8200, DOI 917 10.17487/RFC8200, July 2017, . 920 [RFC5226] Narten, T. and H. Alvestrand, "Guidelines for Writing an 921 IANA Considerations Section in RFCs", RFC 5226, DOI 922 10.17487/RFC5226, May 2008, . 925 9.2 Informative References 927 [RFC7605] Touch, J., "Recommendations on Using Assigned Transport 928 Port Numbers", BCP 165, RFC 7605, DOI 10.17487/RFC7605, 929 August 2015, . 931 [RFC7872] Got, F., Linkova, J., Chown, T., and W. Liu, "Observations 932 on the Dropping of Packets with IPv6 Extension Headers in 933 the Real World", RFC 7872, DOI 10.17487/RFC7872, June 2016, 934 . 936 [TLSCERT] United States Computer Emergency Readiness Team (US-CERT), 937 "Alert (TA17-075A), HTTPS Interception Weakens TLS 938 Security, March 2017 940 [SPUD] Hildebrand, J. and Trammell, B., "Substrate Protocol for 941 User Datagrams (SPUD) Prototype", draft-hildebrand-spud- 942 prototype-03, March 2015 944 [PLUS] Trammell, B. and Kuehlewind, M., "Path Layer UDP Substrate 945 Specification", draft-trammell-plus-spec-01, March 2017 947 [PAN] Trammell, B., "Open Questions in Path Aware Networking", 948 draft-trammell-panrg-questions-02, December 2017 950 [ICMPEH] Herbert, T., "ICMPv6 errors for discarding packets due to 951 processing limits", draft-herbert-6man-icmp-limits-00.txt 953 Author's Address 955 Tom Herbert 956 Quantonium 957 Santa Clara, CA 958 USA 960 Email: tom@quantonium.net