General (Networking) Working Group Internet Draft T. Glassey Document: draft-glassey-dns-rzp-00.txt ServerWerks Inc. Expires: 00/2003 June 2002 ROOT ZONE PROTOCOL draft-glassey-dns-rzp-00.txt Some thoughts on a proposed change to DNS Status of this Memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC2026 [ ]. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. 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." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. Abstract The changing structure and size of today's Internet has taxed the existing name services and their architecture to the breaking point. The DNS system of today was unfortunately architected to have a single root zone and limited set of Top Level Domains. This is defined usually in the set of root servers any resolving request uses to lookup an address. This proposal then is an attempt to lessen the impact on end-users and registrars and to allow IP owners a greater freedom in representing namespace to their customers. The elevator description is that this I-D proposes the restructuring of domain names more fully along the lines of a telephone number by creating a ROOT ZONE PROTOCOL as an addition to multiply the existing DNS services times the number of ROOT ZONES. Table of Contents 1. Document Scope 2 1.1 Document Audience 3 2. Conventions used in this document 3 2.1 Key Words 3 2.2 Special terms 3 2.3 DNS ROOT Servers 4 3. Goals of this I-D 5 4. Existing DNS constraints ū the starting point. 5 4.1 Domain Names, IP and the UDRP Issues 5 4.2 RFC2826 and its limitations 6 4.3 Need to service a more compartmentalized Internet 6 4.4 The limitations of existing naming conventions 7 5. The New proposal ū ROOT ZONE Extended DNS Services. 7 5.1 LQDN based processing 8 5.2 GQDN based processing 8 6. Software Affected by the proposed changes 8 6.1 End-User Clients and their local client-side Resolvers 8 6.2 DNS Servers and their local server-side Resolvers 9 6.3 How this all fits togetherą 10 7. Interim Operations 10 8. Security Considerations 11 9. References 11 10. Acknowledgments 12 11. Author's Addresses 12 1. Document Scope This document addresses a method or reducing the Intellectual property 'collisions', the confusion, pain and suffering with allocating names in DNS land, and does so by creating a discovery protocol for what has until now been a static part of the DNS infrastructure. The DNS Root Zone Tables. The Intent of the Root Zone Protocol (RZP) is to allow for the dynamic mapping of DNS Root Zones at the DNS Server Level and it identifies most all the components that the new Syntax Model would affect as well. This document is a very high level document and will require a number of iterations and the support of other documents to implement its proposed technology including but not limited to a modified version of RFC1034 and 3538 to support its changes at the very least, and likely some changes to the email world to make use of these capabilities. 1.1 Document Audience I assume you know the history of Names and Hostfile formats as per the original works starting at RFC608 and proceeding onward from there. I also assume you know Mockapetris' work on the basic DNS RFC's (starting with RFC1034 and others) and their evolution through what is in place today. I also assume that you understand how zone management works and how DNS is globally administered by the Internet Registrars and how the current Zone Tables are managed OOB. 2. Conventions used in this document 2.1 Key Words 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 RFC-2119 [ ]. 2.2 Special terms The following TERMS are coined in this Document: 2.2.1 ZONE's One specific change is the difference in the use of the term ZONE relative to RFC-2136's use. Their use of the term ZONE refers to which zone within any given root hierarchy. Our use of the term ZONE here is particular to which the Root Hierarchy is being used to resolve any domain name in question. Zones at the use level are represented by Root Tags. The Root tag may be a string of up to 24 characters allowing for the creation of 26 ^^ 24 instances of possible zone names. 2.2.2 ZONE Publication All registrar's supporting Multi-Root operations will make available online an accurate representation of the ROOT ZONES they publish domains in. For each ZONE they add, the registrar and the registrars alone will be responsible for maintaining the accuracy of the Internet's Zone Mapping table(s). Further, this ROOT ZONE publication table will be made available in whatever form the implementation of this protocol extensions demands including through active LDAP lookup or just a simple recursive DNS type model. 2.2.3 Globally Qualified Domain Name (GQDN) The GQDN is a "root zone" enhanced notation model using what was known as a Fully Qualified Domain Name and adding the root zone tag as a prefix to the domain name. 2.2.4 Locally Qualified Domain Name (LQDN) The Locally Qualified Domain Name is identical to what we were calling the Fully Qualified Domain Name. It is called locally qualified since it is assumed to be a part of the system's default root zone. 2.2.5 Registrars ZONE TABLES A locally produced DB.CACHE file and listing of all the ZONES this registrar resolves for. 2.2.6 Zone-Lookup The Zone Lookup is the process of querying the master ZONE TABLE SERVER or a Registrar's local ZONE TABLES for an associated set of ROOT ZONES for any supported set of ZONES. The Operating Authority and all the ZONE-Enabled Registrars maintain ZONE Tables. 2.2.7 Zone-Failure The Failure of the ZONE SERVER to have the correct lookup tables for the requested ZONE(s). It is Zone-Failure that causes the local server to start a ZONE RESOLUTION transaction with the master ZONE TABLE Server. 2.3 DNS ROOT Servers This list of servers defines all of the representations and who will resolve them for this "DNS Hierarchy". Traditionally this is distributed as the DB.CACHE file that BIND uses at boot time. Other Named implementations place it in various other places including as a part of the Active Directory in Microsoft Land. 2.3.1 Single DNS Root This refers to the first implementation rounds at the DNS protocol. It was extended to conform to any number of lookup and record transmittal models all based on the defined hostname and Domain Name structures. In the original DNS service models there is only one root zone and it defines the scope of what DNS can and can't define. 2.3.2 Root Zone Database (RZDB) The RZDB is a local database maintained by the DNS Server regarding which roots it knows how to resolve addresses from. The ROOT ZONE DB or RZDB is a policy centric cache of locally stored DNS ROOT SERVER lists. The Lists are stored locally in textual form and are flushed on a DNS Policy Timeout basis as defined in the ROOT-CACHE-LIFE parameter being added to the Config file (named.boot). 3. Goals of this I-D The goal of this I-D is to propose a new system for the dynamic exchange of ROOT SERVER Data and thus the opening for the ICANN Registrars to support any number of roots. We clearly acknowledge that there are other ways to accomplish aspects of what the Root Zone Protocol enables, but not all of the features can be reproduced through other methods so it is considered a real potential This new system and nomenclature for representing DNS addresses in textual form with the addition of a ZONE or "Root Server" resolution protocol will increase the number of available DOMAIN NAMES in a common format such that any Domain Name can take on a more simple and easier to understand format. The net effect of this will be the increasing of the potential number of ROOT ZONES, ad infinitum. This author feels that this is the best way to meet the commercial expansion of the Internet. 4. Existing DNS constraints ū the starting point. Existing DNS services are architected around a common root set of name servers. These common servers act as an anchor, but specifically limits the root instance to that of the original ARPA. Root's. Today's ROOT Server Lists are maintained by the various Registrars and distributed to their clients out of band. This means that at this time, for the totality of the public Internet, that there is only one of any given name and that the names are parsed from the bottom up (See RFC608. RFC810, and RFC1033/1034, RFC2308, RFC2826 and RFC3258 for more detail). 4.1 Domain Names, IP and the UDRP Issues This facility it is felt will greatly relieve the IP Congestion that today's "One and Only Naming Convention" put in place. This document acknowledges that there are Intellectual Property (IP) similarities between textual, network addresses and the absolute or ASCII representation of the text inside a registered trade or service mark, but that at this time (06/2002), the law surrounding this is still somewhat vague and untested, and as such that the ICANN's UDRP (Uniform domain Dispute Resolution Protocol) was developed to protect ICANN and the registrars from IP suits. To date the UDRP hasn't worked to well, but much of the need for it is based in the "there can only be one of any network name" and for the most part this is based on that there is only one dot com, net or orgą We hope to address that. The other issue is which parts of the domain names are allowed by design to carry any customer definable content, and that is at the Second Level Domain Name (SLDN) only. So this further restricts the possible combinations from "phrase style identity tags" to the current resolving structures. 4.2 RFC2826 and its limitations THE IAB published a document in May of 2000 complaining of the problems of addressing the dynamic ROOT management of the Internet. Thy cited three cause why this was not doable today, and these were: 1) The need to maintain a common naming and nomenclature standard 2) The complexity of the NAME ZONE update processes 3) And finally that traditional ROOT TABLES were poorly maintained and were distributed in the existing DNS mode out-of-band (OOB). None of these are insurmountable problems and in fact the second one has no bearing on the issues of ROOT ZONE MANAGEMENT. 4.3 Need to service a more compartmentalized Internet Nowadays, and for whatever reason, the Internet is becoming more and more compartmentalized. This may be due to concepts like eBorders and jurisdictional dilemmas being resolved in this manner, or that the economics of the Internet are directly tied to the economic outlook of the world and that since the slowdown of last year and 9/11 that drastic changes in what the Internet physically consists of are now in play. Even the Internet Architecture Board in their report of May 2000 criticized the existence of only a single DNS root [rfc-2826]. 4.4 The limitations of existing naming conventions The real issue with textual name space is how small the usable portion of it is. There are two key constraints defining the scope of the namespace that is relevant in the current model and that is the totality of the second level domain names. There is no way for the Domain Client to define the Top Level domains, and since these are the same for all Registrars pretty much, this means that the totality for the space they can sell is that of the 2nd level domains. Inside of this Second Level Domain envelope is the further limitation of that relatively very few of the possible combinations of letters actually form words or symbols that are identifiable, further constricting the totality of the number of realistically representable names that and TLD can support. And when you compound that with existing domain name needs, there is a meltdown coming of available names. 5. The New proposal ū ROOT ZONE Extended DNS Services. This new proposal for bringing ROOT ZONE MANAGEMENT under the scope of what is provided from common DNS Services, and through the dynamic opening of ROOT TAG SPACE, will address the creation of a more wide-open namespace. The new DNS Nomenclature proposal is based around the addition of a ROOT ZONE TAG or RZT's as a zone identifier code, which is prepended like an Area Code, to the front of the existing FQDN. This system supports the existing FQDN name resolution, but as the default set of ROOT SERVERS that the DNS SERVER uses to resolve queries to it. The Globally Qualified Domain Name (GQDN) will be constructed as follows: <[ROOT ZONE TAG]>. And the LQDN (a.k.a. the FQDN) is deconstructed as follows: ... o- Where [ROOT ZONE TAG] is the optional bracket enclosed text string indicating which group of ROOT ZONE servers to use for this query. o- Where Hostname = any reasonable RFC compliant Hostname o- Where Second Level Domain = the existing settable or user definable portion of the FQDN, and; o- Where the final root of the name is the anchoring TLD or DNS Zone Table 5.1 LQDN based processing The LQDN structure is parsed and processed exactly as the FQDN is today. The name is sent to the NAME SERVER for resolution and the name server's parser will see that it is a LQDN and so this name server will use its Default Set of ROOT SERVERS. 5.2 GQDN based processing If a ROOT ZONE TAG (RTZ) is prepended onto the Host name, the DNS Server will process that tag to see if it is either the default tag name or that of an already known root zone. For any root zone it already "knows" there will be a local copy of that root zone in the DNS Server's ROOT ZONE Database. 6. Software Affected by the proposed changes It is important to understand that this change and new requirement may break some existing resolver/parser code since they may be hardwired to the existing domain structure. But any good programmer knows that it is easy to create a secondary footprint and set of parsing rules such that if there is a ROOT ZONE TAG (RZT) detected in the host's domain address specification, that it would be resolvable. The following Services are expected to require some work to address this new set of ZONE management facilities. 6.1 End-User Clients and their local client-side Resolvers Depending on how much parsing the client side applications do, there may be some trivial extending of the parsers to accept the bracket delimited "[ROOT ZONE TAG]" as the prepended extension to the LQDN's address. But in many instances it is more likely that the OS's resolver is the totality of what the application uses and the apps just pass text strings to the local resolver input. Therefore extending it to support the "[ROOT ZONE TAG]" extensions is not such a big deal, especially since it is easily cut in as an optional feature in most all Resolvers. 6.2 DNS Servers and their local server-side Resolvers DNS Resolvers themselves may be there largest part of the recoding effort, but this next generation syntax support can be rolled within a major BIND release, so this will not be such a painful effort. Once at least one generation of RZP enhanced DNS is released the rest of the infrastructure will be deployed as well. 6.2.1 Changing sets of Root Servers dynamically The key issues with the DNS Servers that needs to be upgraded to support root ZONE management is the process within a running DNS server of changing its root pointers. Most DNS Servers' root pointers are currently read from static files as text strings, and in most instances, at least UNIX ones, there is a need to create a dynamic set of pointer indirections that the actual values pointed to by the ROOT SERVER Template, can be mapped in and out of at will (and without a hang-up or HUP signal). This is easily accomplished by creating a dummy list, or template as a global and then loading the named selected ROOT SERVER Entries into it at will. 6.2.2 Need to add the Root Zone Database (RZDB)ą and it's processing. Somehow the local server needs to create a ROOT CACHE for each ROOT ZONE it is to resolve for. These zones can remain pretty static or expire depending on who and what. The only real reasons that they would change once downloaded would be that registrar was significantly changing their service topology or going out of business. So its no big deal to cache the zone tables at any DNS server for 6 months at a time or more; 6.2.3 Other TOOLS for DNS management - Dig, nslookup, etc. This is the real heartbreaker. For tools that do not rely on the OS's underlying resolver but supply their own, there is a significant amount of recoding here at the toolsmith level to deal with the expanded namespace and ROOT ZONE resolution. Oh well, that's life is my only response here. We as a culture need the ROOT SPACE too much to not suffer whatever pain this causes. 6.3 How this all fits togetherą The Use Modelsą Client requests the resolving of a GQDN's address from the Name Server. The server parses the ROOT ZONE TAG (RZT) from the GQGN and looks up in its local ROOT ZONE TAG Cache to see if the TAG and its equates exist there. If so - Then the server checks the expiry of the equate list to make sure it can use the entry. If the expiry is still good then it resolves the name and passes the address back to the client's Application Otherwise ZONE FAIL occurs. ROOT ZONE FAILs can be addressed through policy in a number of manners, and as a default mode, we propose that the server attempts to recover a new copy of the ZONE TABLE from the master server or from the previous supplier of this ZONE TABLE, which was also cached locally in the Server. If the ZONE recovery is completed then the DNS Server caches the updated ZONE Table entry, updates its ZONE EXPIRY and continues to process the name resolution. To continue, with the new ZONE TABLE's list of root servers, the DNS Server looks up the address and assuming a successful response, sends the address onward into the client. If there is a lookup failure then the user is told that there was a standard DNS resolution error, or the process times out and the client barfs appropriately, just as with today's SINGLE ROOT service model. If the ZONE TABLE transfer fails then the DNS server has the choice of continuing to use the old table if it works and continuing the attempt to resolve the name. 7. Interim Operations The problem with Interim operations is that there may be name collisions between ROOT 'a' and ROOT 'b' such that addresses that exist in both are never reached in the second ROOT because those in the first root will take precedence. To this end we see ROOT ZONES and the fact that there are now several mutually exclusive ones up and functioning on the Internet as a serious problem without this bridge process to allow for selective ROOT ZONE management. 8. Security Considerations There are no real security considerations with this change or augmentation of existing DNS Services. 9. References [RFC-830] Z. Su, "A Distributed System for Internet Name Service", IETF RFC-830, Network Information Center, SRI International, October 1982. Contains early thoughts on the design of the domain system. [RFC-882] P. Mockapetris, "Domain names - Concepts and Facilities," RFC-882, USC/Information Sciences Institute, November 1983. [RFC-883] P. Mockapetris, "Domain names - Implementation and Specification," RFC-883, USC/Information Sciences Institute, November 1983. [RFC-920] J. Postel and J. Reynolds, "Domain Requirements", RFC-920, USC/Information Sciences Institute, October 84. [RFC-952] K. Harrenstien, M. Stahl, E. Feinler, "DoD Internet Host Table Specification", RFC-952, SRI, October 85. [RFC -1033] M. Lottor, DOMAIN ADMINISTRATORS OPERATIONS GUIDE SRI, International, November 1987 [RFC-1034} P. Mockapetris DOMAIN NAMES - CONCEPTS AND FACILITIES, ISI November 1987 [RFC 2136] Vixie, Et al., DNS Update, April 1997 [RFC 2308] M. Andrews, DNS NCACHE (Negative Caching of DNS), March 1998 [RFC-2782] A. Gulbrandsen, P. Vixie, L. Esibov, DNS SRV Resource Records, February 2000 [RFC-2826] IAB Technical Comment on the Unique DNS Root, May 2000 [RFC-3258] Distributing Authoritative Name Servers April 2002 10. Acknowledgments Vint Cerf ū For taking all the crap I have dished out to him for all these years. My friends in the IETF including those PKIX and POISSION WG Chairs and its Secretariat who I also gave merde to for so long. Those inventors of DNS and the Host Naming Conventions named in the references section of this document and those critical Telephone Pioneers (John Draper and others). 11. Author's Addresses Todd S. Glassey ServerWerks PO Box 0026, Menlo Park, Ca., 94026 Phone: 650-926-9609 Email: Todd.Glassey@ServerWerks.CC Root Zone Protocol June 2002 GLASSEY Expires December 2002