RFC 2208

Resource ReSerVation Protocol (RSVP) Version 1 Applicability Statement: Some Guidelines on Deployment ✨

Okay bestie, let's get into it 🎀 This is RFC 2208, a document from the Network Working Group published in September 1997. It's marked Category: Informational, so it's guidance, not a standard. The author list is A. Mankin, Ed. (USC/ISI), F. Baker (Cisco Systems), B. Braden (USC/ISI), S. Bradner (Harvard), M. O`Dell (UUNET Technologies), A. Romanow (Sun Microsystems), A. Weinrib (Intel Corporation), and L. Zhang (UCLA) 👑. The header doesn't list anything this RFC obsoletes or updates.

Status of this Memo 📝

This memo gives information to the Internet community 💖 It does not specify an Internet standard of any kind, and distribution of it is unlimited 🦋.

Abstract 🌸

Okay so this doc explains where RSVP fits alongside the Integrated Services protocols and the other pieces of resource reservation, and gives deployment tips for right now 🧚. It's also the companion for the first submission of RSVP and the integrated services specs onto the Internet standards track, so it's kind of a big deal 🔥.

1. Introduction ✨

RSVP [RFC 2205] is a unicast and multicast signalling protocol 🍓, and its whole job is to install and maintain reservation state at each router along the path of a data stream 💅. The state it handles is defined by the services in [RFC 2211] and [RFC 2212], which came out of the Integrated Services WG 🦋. RSVP and those services are being introduced to the IETF standards track together, so from here on, when we just say "RSVP," we mean the signalling protocol plus the integrated service specs 🫶.

RSVP needs several extra components to work, and they're listed in Table 1 👑:

Table 1  Additional Components of Resource Reservation

1. Message formats for expressing the parameters of the services you want. A proposed standard set of these formats is specified in [RFC 2210].
2. Router and host mechanisms (for example, packet classification and scheduling, and admission control algorithms) that implement one or both of the models in [RFC 2211] and [RFC 2212], which are also on the standards track ✨.
3. Message formats for expressing policies about admission control and resource use. A small common subset of these formats for the standards track is in the RSVP WG's charter. The Policy objects in the RSVP Protocol Specification are optional only at this time 🙈.
4. Mechanisms spread across the network that carry out admission control policy functions, including authorization and other security mechanisms 🔐.

When you have some form of each component in Table 1, RSVP-enabled apps can get differentiated qualities of service across IP networks 🌈. Networked multimedia apps are the obvious early users, since many of them need (or would benefit from) a predictable end-user experience 🎀.

RSVP and the integrated services and other components in Table 1 are a big change to the service model of IP networks, so RSVP got lots of attention and press before it was even a standards track RFC 📰. Lots of OS and router vendors are building RSVP and integrated services into their products for near-future release 💀. The goal of this applicability statement is to describe the uses of the current RSVP spec that are known to be feasible, and to flag the limitations along with the ongoing chartered work that's tackling them 🧚.

2. Issues Affecting Deployment of RSVP ⚠️

Rolling out RSVP at scale needs real care, because there are still some open issues that could affect whether the deployment succeeds 🙈.

2.1. Scalability 📈

The resource requirements (processing and storage) for running RSVP on a router grow in proportion to the number of separate sessions, meaning RSVP reservations 💅. So supporting lots of small reservations on a high-bandwidth link can easily overload the routers, and that's a bad idea 💀. On top of that, the packet classification and scheduling that currently provide differentiated services for reserved flows may be really hard for some router products, or on some of their high-speed interfaces (e.g. OC-3 and above) 🔥.

These scaling issues mean it generally won't be appropriate to deploy RSVP on high-bandwidth backbones right now 🌸. Looking ahead, backbone operators probably won't go with naively implementing RSVP for each separate stream 🙅. Instead, techniques are being developed that aggregate the streams needing special treatment at the "edge" of the backbone 🦋. Inside the backbone, less costly approaches would then set aside resources for the aggregate as a whole, as a way to meet the end-to-end requirements of individual flows ✨.

In the near term, different vendors will probably take different approaches to aggregating reservations 🌈. There isn't any chartered IETF work on developing standards in this space yet 😭. A BOF called Future Directions of Differential Services, held on April 7, 1997, at the Memphis IETF, was going to look at the IETF's next steps here, among other issues 💜. Public documentation of aggregation techniques and experience is encouraged 🫶.

2.2. Security Considerations 🔐

Okay, this part is serious, so let's be precise 💅. The RSVP WG submission for Proposed Standard includes two security-related documents, [Baker96, RFC 2207] 📚. [Baker96] deals with denial and hijacking or theft of service attacks 💀. [RFC 2207] covers RSVP mechanisms for data flows that themselves use IPSEC 🔒.

The first document is meant to protect against spoofed reservation requests arriving at RSVP routers 🚨. Those requests could be used to get service for unauthorized parties, or to lock up network resources in a denial of service attack 😱. Modified and spoofed reservation requests are detected with hop-by-hop MD5 checksums (in an Integrity Object) between RSVP neighbor routers 🔍. As described, RSVP hop-by-hop authentication assumes that key management and distribution for routers is already sorted and deployed 🗝️. Until an effective key infrastructure is in place, manually keyed session integrity might be used instead 🛠️. Also, [Baker96] may be updated with RFC 2085 🌸.

Here's the thing: RSVP needing an effective key infrastructure among routers isn't unique to RSVP 🙄. It's widely acknowledged that there are plenty of denial of service attacks on the routing infrastructure (totally independent of RSVP), and those will only be resolved by deploying a key infrastructure 💀.

Theft of service risks mean you have to deploy with caution, bestie 🚨. A basic precaution is to configure management logging of new and changed filter specifications in RSVP-enabled infrastructure, for example the newFlow trap [RFC 2206] 📋.

The Integrity object defined by [Baker96] may also play a role in policy control, which will be covered in 2.3 🎀.

The second security-related document gives a mechanism for carrying flows in which the transport and user octets have been encrypted (RFC 1827) 🔐. Encryption like that is hugely beneficial for certain apps, but it's not relevant to the functional security of RSVP or of reservations 🙅.

The next section on Policy Control has more on RSVP authorization security 👑.

2.3. Policy Control 🧚

Policy control is all about who can, or can't, make reservations once a reservation protocol can be used to set up unequal services 💅.

Right now, the RSVP spec defines a mechanism for carrying policy information along with reservations 🎀. But the spec doesn't define the policies themselves 🙈. For now, vendors have said they'll use the RSVP-defined mechanism to implement proprietary policies 🏷️.

The RSVP WG is chartered to specify a simple standardized policy object, plus complete simple mechanisms for session use of the Integrity object, in the near future ✨. This applicability statement may be updated when the WG's charter is complete 🌈.

Before deciding to deploy RSVP, it would be wise to make sure the policy control a vendor offers is good enough for what you intend to use it for 🧐. On top of the lack of documented policy mechanisms in any policy area (like access control, authorization, and accounting), the community has very little experience with describing, setting, and controlling policies that limit Internet service 😬. So vendor solutions will probably get revised a lot, especially before the IETF has come up with any policy specification 🔄.

---

That's part 1, bestie 🌈💖

4. References ✨

okay so the references are literally the receipts for everything this doc leans on 🧾 here's the whole list, no cap:

• [Baker96] Baker, F., "RSVP Cryptographic Authentication", Work in Progress.
• [RFC 2206], Baker, F. and J. Krawczyk, "RSVP Management Information Base", RFC 2206, September 1997.
• [RFC 2207] Berger, L. and T. O'Malley, "RSVP Extensions for IPSEC Data Flows", RFC 2207, September 1997.
• [RFC 2211] Wroclawski, J., "Specification of Controlled-Load Network Element Service", RFC 2211, September 1997.
• [RFC 2212] Shenker, S., C. Partridge and R. Guerin, "Specification of Guaranteed Quality of Service", RFC 2212, September 1997.
• [RFC 2205] Braden, R. Ed. et al, "Resource ReserVation Protocol -- Version 1 Functional Specification", RFC 2205, September 1997.
• [RFC 2210] Wroclawski, J., "The Use of RSVP with IETF Integrated Services", RFC 2210, September 1997.

---

5. Authors' Addresses 💅

the whole squad who wrote this one 🫶 contact info is right here, exactly as it was written:

Fred Baker                    Abel Weinrib
Cisco Systems                 Intel Corporation
Phone: 408-526-4257           Phone: 503-264-8972
EMail: fred@cisco.com         EMail: aweinrib@ibeam.intel.com

Bob Braden
USC/ISI                       Lixia Zhang
4676 Admiralty Way            UCLA Computer Science Department
Marina Del Rey, CA 90292      4531G Boelter Hall
Phone: 310-822-1511           Los Angeles, CA 90095-1596 USA
EMail: braden@isi.edu         Phone: 310-825-2695
                              EMail: lixia@cs.ucla.edu

Scott Bradner                 Allyn Romanow
Harvard University            Sun Microsystems
Phone: 617-495-3864           Phone: 415-786-5179
EMail: sob@harvard.edu        EMail: allyn@eng.sun.com

Michael O'Dell         Allison Mankin
UUNET Technologies         mankin@east.isi.edu
Phone: 703-206-5471
EMail: mo@uu.net

that's the whole RFC, bestie 🌈🦋 the protocol nerds really said "let's make RSVP work" and then wrote it all down so we could all slay with it 👑