ENUM L. Conroy Internet-Draft RMRL Expires: December 27, 2006 K. Fujiwara JPRS June 25, 2006 ENUM Implementation Issues and Experiences Status of this Memo By submitting this Internet-Draft, each author represents that any applicable patent or other IPR claims of which he or she is aware have been or will be disclosed, and any of which he or she becomes aware will be disclosed, in accordance with Section 6 of BCP 79. 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. This Internet-Draft will expire on December 27, 2006. Copyright Notice Copyright (C) The Internet Society (2006). Abstract This document captures experience in implementing systems based on the ENUM protocol, and experience of ENUM data that have been created by others. As such, it is advisory, and produced as a help to others in reporting what is "out there" and the potential pitfalls in interpreting the set of documents that specify the protocol. Conroy & Fujiwara Expires December 27, 2006 [Page 1] Internet-Draft ENUM Experiences June 2006 Table of Contents 1. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3 2. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.1. Interpretation of Recommendations . . . . . . . . . . . . 4 3. Character Sets and ENUM . . . . . . . . . . . . . . . . . . . 6 3.1. Character Sets - Non-ASCII considered harmful . . . . . . 6 3.2. Case Sensitivity . . . . . . . . . . . . . . . . . . . . . 9 3.3. RegExp field delimiter . . . . . . . . . . . . . . . . . . 9 3.4. RegExp Meta-character Issue . . . . . . . . . . . . . . . 10 4. ORDER/PRIORITY Processing . . . . . . . . . . . . . . . . . . 11 4.1. Order/Priority values - general processing . . . . . . . . 11 4.2. NAPTRs with identical ORDER/PRIORITY values . . . . . . . 13 4.2.1. Compound NAPTRs and implicit ORDER/REFERENCE Values . 13 4.3. Compound NAPTR Processing . . . . . . . . . . . . . . . . 14 4.4. Processing Order value across Zones . . . . . . . . . . . 14 5. Non-Terminal NAPTR Processing . . . . . . . . . . . . . . . . 16 5.1. Non-Terminal NAPTRs - necessity . . . . . . . . . . . . . 16 5.2. Non-Terminal NAPTRs - future implementation . . . . . . . 17 5.2.1. Non-Terminal NAPTRs - general . . . . . . . . . . . . 17 5.2.2. Non-Terminal NAPTRs - loop detection and response . . 17 5.3. Interpretation of RFC 3403 and RFC 3761 . . . . . . . . . 18 5.3.1. Flags field content with Non-Terminal NAPTRs . . . . . 18 5.3.2. Services field content with Non-Terminal NAPTRs . . . 18 5.3.3. Regular Expression and Replacement field content with non-terminal NAPTRs . . . . . . . . . . . . . . . 19 6. General DNS Issues . . . . . . . . . . . . . . . . . . . . . . 22 6.1. DNS Specifications . . . . . . . . . . . . . . . . . . . . 22 6.2. ENUM needs EDNS0 support . . . . . . . . . . . . . . . . . 22 6.3. ENUM EDNS0 message size support . . . . . . . . . . . . . 24 6.4. Intermediary Devices . . . . . . . . . . . . . . . . . . . 25 6.5. Times To Live and NAPTRs . . . . . . . . . . . . . . . . . 26 7. Backwards Compatibility . . . . . . . . . . . . . . . . . . . 28 7.1. Services field syntax . . . . . . . . . . . . . . . . . . 28 8. Security Considerations . . . . . . . . . . . . . . . . . . . 30 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 31 10. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 32 11. References . . . . . . . . . . . . . . . . . . . . . . . . . . 33 11.1. Normative References . . . . . . . . . . . . . . . . . . . 33 11.2. Informative References . . . . . . . . . . . . . . . . . . 34 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 35 Intellectual Property and Copyright Statements . . . . . . . . . . 36 Conroy & Fujiwara Expires December 27, 2006 [Page 2] Internet-Draft ENUM Experiences June 2006 1. Terminology This document is Advisory, and does not specify a standard of any kind. Note that recommendations here contain the words "MUST", "REQUIRE", "SHOULD", and "MAY". This particular document does not form a standard and so these terms DO NOT hold their normative definitions. The proposals include these terms from observation of behaviour and for internal consistency, where Client and Server recommendations have to match. Conroy & Fujiwara Expires December 27, 2006 [Page 3] Internet-Draft ENUM Experiences June 2006 2. Introduction The ENUM protocol ([1]) and the Dynamic Delegation Discovery System (DDDS, [2] [3] [4] [5] [6]) are defined elsewhere, and those documents alone form the normative definition of the ENUM system. Unfortunately, this document cannot provide an overview of the specifications, so the reader is assumed to have read and understood the complete set of ENUM normative documents. From experience of creating ENUM data and of developing client systems to process that data it is apparent that there are some subtleties in the specifications that have led to different interpretations; in addition there are common syntactic mistakes in data currently "out there" on the Internet. This document is intended to help others avoid the potential pitfalls in interpreting the set of documents that specify the protocol. It also reports the kind of data they will "find" and so how to process the intent of the publisher of that ENUM data, regardless of the syntax used. As such, it is in keeping with the principle evinced in RFC 791 that "In general, an implementation must be conservative in its sending behavior, and liberal in its receiving behavior". Of course, no client should fail if it receives data that is illegal; typical "defensive programming" techniques should be applied with ENUM clients as with any other protocol handler. DNS processing does add some factors not typical for other protocols. For example, it is legal (but unwise) for a string within a NAPTR to include a NUL character (U+0000). Such a character can even legally be used as a delimiter in the RegExp field. This may well cause problems for many clients that rely on simplistic 'C' string processing libraries. Whilst we advise specifically against that (see Section 3.3), and more generally, use of "non-printable" characters (see Section 3.1), a client may encounter these and must not assume that everyone shares the same assumptions on design or data validity; a string has a very specific meaning in DNS. Finally, note that the DDDS system is intricate and so in some places there are several potential interpretations of the specifications. This document proposes a suggested interpretation for some of these points, but they are just that; suggestions. 2.1. Interpretation of Recommendations Any ENUM implementation issue has two sides: o the "Server" side covering the expected behaviour of the ENUM zone provisioning system and expectations Registrants may make, and Conroy & Fujiwara Expires December 27, 2006 [Page 4] Internet-Draft ENUM Experiences June 2006 o the "Client" side covering behaviour that has been observed and that can be expected of the Client, together with the expectations that an end user who requests an ENUM lookup may make. For each of the issues, we have split the recommendations into "Client" and "Server" proposals. In three cases, we have indicated proposals that relate to ENUMservice specifications, rather than implementations; these are labelled as "Spec". Also, there is one recommendation that concerns "Middleboxes" (such as any intervening Firewalls) rather than the DNS entities involved directly in an ENUM query. This recommendation is labelled as "MidBox". Few if any Client programs handling ENUM NAPTRs have implemented all of the features described in RFC 3761 and in the DDDS specifications. Thus the recommendations have tried to follow best practice rather than leaving interpretation to each of the individual implementation teams. Likewise, we have reflected the likely behaviour of Client programs, rather than being purely prescriptive. Any developer of Provisioning Systems that populate ENUM NAPTRs will need to know what his or her "opposite number" developing Client programs has decided - to do anything else would be a disservice. There are undoubtedly other issues, and developers are asked to raise any others they find on the appropriate IETF Working group's mailing list and/or by e-mail to the authors (see later for contact information). Finally, note that the authors are not aware of any IPR issues that are involved in the suggestions made in this document. Conroy & Fujiwara Expires December 27, 2006 [Page 5] Internet-Draft ENUM Experiences June 2006 3. Character Sets and ENUM 3.1. Character Sets - Non-ASCII considered harmful RFC 3761 and RFC 3403 ([1] and [2]) specify respectively that ENUM and NAPTRs support Unicode using the UTF-8 encoding specified in [7]. This raises an issue where implementations use "single byte" string processing routines. If there are multi-byte characters within an ENUM NAPTR, incorrect processing may well result from these "UTF-8 unaware" systems. The UTF-8 encoding has a "US-ASCII equivalent range", so that all characters in US-ASCII [20] from 0x00 to 0x7F hexadecimal have an identity map to the UTF-8 encoding; the encodings are the same. In UTF-8, characters with Unicode code points above this range will be encoded using more than one byte, all of which will be in the range 0x80 to 0xFF hexadecimal. Thus it is important to consider the different fields of a NAPTR and whether or not multi-byte characters can or should appear in them. In addition, characters in the "non-printable" portion of US-ASCII (0x00 to 0x1F hexadecimal, plus 0x7F hexadecimal) are "difficult". Although NAPTRs are processed by machine, they may sometimes need to be written in a "human readable" form. Similarly, if NAPTR content is shown to an end user so that he or she may choose, it is important that the content is "human readable". Thus it is unwise to use non- printable characters within the US-ASCII range; the ENUM client may have good reason to reject NAPTRs that include these characters as they cannot readily be presented to an end-user. There are two numeric fields in a NAPTR; the ORDER and PREFERENCE/ PRIORITY fields. As these contain binary values, no risk is involved as string processing should not be applied to them. The "string based" fields are the Flags, Services, and RegExp fields. The Replacement field holds a domain name encoded according to the standard DNS mechanism [8][9]. With the introduction of Internationalized Domain Name (IDN) support, this domain name MUST be further encoded using Punycode [10]. As this holds a domain name that is not subject to replacement or modification (other than Punycode processing), it is not of concern here. Taking the string fields in turn, the Flags field contains characters that indicate the disposition of the NAPTR. This may be empty, in which case the NAPTR is "non-terminal", or it may include a flag character as specified in RFC 3761. These characters all fall into the US-ASCII equivalent range, so multi-byte characters cannot occur. The Services field includes the DDDS Application identifier ("E2U") Conroy & Fujiwara Expires December 27, 2006 [Page 6] Internet-Draft ENUM Experiences June 2006 used for ENUM, the '+' character used to separate tokens, and a set of ENUMservice identifiers, any of which may include the ':' separator character. In section 2.4.2 of RFC 3761 these identifiers are specified as 1*32 ALPHA/DIGIT, so there is no possibility of non- ASCII characters in the Services field. The RegExp field is more complex. It forms a sed-like substitution expression, defined in [2], and consists of two sub-fields: o the POSIX Extended Regular Expression (ERE) sub-field [11] o a replacement (repl) sub-field [2]. Additionally, RFC 3403 specifies that a flag character may be appended, but the only flag currently defined there (the 'i' case insensitivity flag) is not appropriate for ENUM - see later in this document. The ERE sub-field matches against the "Application Unique String"; for ENUM, this is defined in RFC 3761 to consist of digit characters, with an initial '+' character. It is similar to a global-number- digits production of a tel: URI, as specified in [12], but with visual-separators removed. In short, it is a telephone number (see [13]) in restricted format. All of these characters fall into the US-ASCII equivalent range of UTF-8 encoding, as do the characters significant to the ERE processing. Thus, for ENUM, there will be no multi-byte characters within this sub-field. The repl sub-field can include a mixture of explicit text used to construct a URI and characters significant to the substitution expression, as defined in RFC 3403. Whilst the latter set all fall into the US-ASCII equivalent range of UTF-8 encoding, this might not be the case for all conceivable text used to construct a URI. Presence of multi-byte characters could complicate URI generation and processing routines. URI generic syntax is defined in [14] as a sequence of characters chosen from a limited subset of the repertoire of US-ASCII characters. The current URIs use the standard URI character "escaping" rules specified in the URI generic syntax, and so any multi-byte characters will be pre-processed; they will not occur in the explicit text used to construct a URI within the repl sub-field. However, the Internationalized Resource Identifier (IRI) is defined in [15] as extending the syntax of URIs, and specifies a mapping from an IRI to a URI. IRI syntax allows characters with multi-byte UTF-8 encoding. Given that this is the only place within an ENUM NAPTR where such Conroy & Fujiwara Expires December 27, 2006 [Page 7] Internet-Draft ENUM Experiences June 2006 multi-byte encodings might reasonably be found, a simple solution is to use the mapping method specified in section 3.1 of [15] to convert any IRI into its equivalent URI. This process consists of two elements; the domain part of an IRI MUST be processed using Punycode if it has a non-ASCII domain name, and the remainder MUST be processed using the extended "escaping" rules specified in the IRI document if it contains characters outside the normal URI repertoire. Using this process, there will be no non- ASCII characters in any part of any URI, even if it has been converted from an IRI that contains such characters. Taking into account the existing client base, it is RECOMMENDED that: Spec ENUMservice registrations should REQUIRE that any static text in the repl sub-field is encoded using only characters in the US-ASCII equivalent range that are "printable". If any of the static text characters do fall outside this range then they MUST be pre-processed using an IRI/ URI-specific "escape" mechanism to re-encode them only using US-ASCII equivalent printable characters (those in the range U+0020 to U+007E). At the least, it is RECOMMENDED that: Spec Any ENUMservice registration that allows characters requiring multi-byte UTF-8 encoding to be present in the repl sub-field must have a clear indication that there may be characters outside of the US-ASCII equivalent range. Such an ENUMservice registration is strongly discouraged, as the mechanisms specified in section 3.1 of [15] will suffice. Finally, the majority of ENUM clients in use today do not support multi-byte encodings of the Unicode Consortium's Universal Character Set (UCS). This is a reasonable choice, particularly for "small footprint" implementations, and they may not be able to support NAPTR content that is non-printable as they need to present the content to an end user for selection. Thus, it is RECOMMENDED that: Client ENUM clients may discard NAPTRs in which they detect characters not in the US-ASCII "printable" range (0x20 to 0x7E hexadecimal). ENUM zone provisioning systems should consider this. It is RECOMMENDED that: Conroy & Fujiwara Expires December 27, 2006 [Page 8] Internet-Draft ENUM Experiences June 2006 Server ENUM zone provisioning systems should not use non-ASCII characters in the NAPTRs they generate unless it is clear that all ENUM clients they are designed to support will be able correctly to process such characters. 3.2. Case Sensitivity The only place where NAPTR field content is case sensitive is in any static text in the repl sub-field of the RegExp field. Everywhere else, case insensitive processing can be used. The case insensitivity flag ('i') could be added at the end of the RegExp field. However, in ENUM, the ERE sub-field operates on a string defined as the '+' character, followed by a sequence of digit characters. Thus this flag is redundant for E2U NAPTRs, as it does not act on the repl sub-field contents. To avoid the confusion that this generates, It is RECOMMENDED that: Server When populating ENUM zones with NAPTRs, provisioning systems should not use the 'i' RegExp field flag, as it has no effect and some ENUM clients do not expect it. Client ENUM clients should not assume that the delimiter is the last character of the field. 3.3. RegExp field delimiter It is not possible to select a delimiter character that cannot appear in one of the sub-fields. Some old clients are "hardwired" to expect the character '!' as a delimiter. This is used in an example in RFC 3403. It is RECOMMENDED that: Server ENUM zone provisioning systems should use '!' (U+0021) as their RegExp delimiter character. Client ENUM clients may discard NAPTRs that do not use '!' as a RegExp delimiter. This character cannot appear in the ERE sub-field. It may appear in the content of some URIs, as it is a valid character (e.g. in http URLs). Thus, it is further RECOMMENDED that: Conroy & Fujiwara Expires December 27, 2006 [Page 9] Internet-Draft ENUM Experiences June 2006 Server ENUM zone provisioning systems must ensure that, if the RegExp delimiter is a character in the static text of the repl sub-field, it must be "escaped" using the escaped- delimiter production of the BNF specification shown in section 3.2 of RFC 3402 (i.e. "\!", U+005C U+0021). Finally, in keeping with RFC 3402: Client ENUM clients should discard NAPTRs that have more or less than 3 "unescaped" instances of the delimiter character within the RegExp field. 3.4. RegExp Meta-character Issue In ENUM, the ERE sub-field may include a literal character '+', as the Application Unique String on which it operates includes this. However, if it is present, then '+' must be "escaped" using a single backslash character as '+' is a meta-character in POSIX Extended Regular Expression syntax. The following NAPTR example is incorrect: * IN NAPTR 100 10 "u" "E2U+sip" "!^+46555(.*)$!sip:\1@example.net!" . This example MUST be written as: * IN NAPTR 100 10 "u" "E2U+sip" "!^\+46555(.*)$!sip:\1@example.net!" . Thus, it is RECOMMENDED that: Server If present in the ERE sub-field of an ENUM NAPTR, '+' must be written as "\+" (i.e. U+005C U+002B). Conroy & Fujiwara Expires December 27, 2006 [Page 10] Internet-Draft ENUM Experiences June 2006 4. ORDER/PRIORITY Processing 4.1. Order/Priority values - general processing RFC 3761 and RFC 3403 state that the ENUM client MUST sort the NAPTRs using the ORDER field value ("lowest value is first") and SHOULD order the NAPTRs using the PREFERENCE/PRIORITY field value as the minor sort term (again, lowest value first). The NAPTRs in the sorted list must be processed in order. Subsequent NAPTRs with less preferred ORDER values must only be dealt with once the current ones with a "winning" ORDER value have been processed. However, this expected behaviour is a simplification; ENUM clients may not behave this way in practice, and so there is a conflict between the specification and practice. For example, ENUM clients will be incapable of using most NAPTRs as they do not support the ENUMservice (and the URI generated by those NAPTRs). As such, they will discard the "unusable" NAPTRs and continue with processing the "next best" NAPTR in the list. The end user may have pre-specified his or her own preference for services to be used. Thus, an end user may specify that he or she would prefer to use contacts with a "sip" ENUMservice, and then those with "email:mailto" service, and is not interested in any other options. Thus the sorted list as proposed by the Registrant (and published via ENUM) may be reordered. For example, a NAPTR with a "sip" ENUMservice may have a "losing" ORDER field value, and yet is chosen before a NAPTR with an "h323" ENUMservice and a "winning" ORDER value. This may occur even if the node the end user controls is capable of handling other ENUMservices. ENUM clients may also include the end user "in the decision loop", offering the end user the choice from a list of possible NAPTRs. Given that the ORDER field value is the major sort term, one would expect a conforming ENUM client to present only those NAPTRs with a "winning" ORDER field value as choices. However, if all the options presented had been rejected, then the ENUM client might offer those with the "next best" ORDER field value, and so on. As this may be confusing for the end user, some clients simply offer all of the available NAPTRs as options to the end user for his or her selection "in one go". In summary, some ENUM clients will take into account the Services field value along with the ORDER and PREFERENCE/PRIORITY field values, and may consider the preferences of the end user. The Registrant and the ENUM zone provisioning system he or she uses must be aware of this and should not rely on ENUM clients taking Conroy & Fujiwara Expires December 27, 2006 [Page 11] Internet-Draft ENUM Experiences June 2006 account of the value of the ORDER and the PREFERENCE/PRIORITY fields. Specifically, it is unsafe to assume that a ENUM client will not consider another NAPTR until it has discarded one with a "winning" ORDER value. The instruction (in RFC 3403 section 4.1 and section 8) may or may not be followed strictly by different ENUM clients for perfectly justifiable reasons. To avoid the risk of variable behaviour, it is RECOMMENDED that: Server ENUM zone provisioning systems should not use different ORDER values for NAPTRs within a zone. In our experience, incorrect ORDER values in ENUM zones is a major source of problems. Although it is by no means required, it is further RECOMMENDED that: Server ENUM zone provisioning systems should use a value of 100 as the default ORDER value to be used with all NAPTRs. As such, when populating a zone with NAPTRs, it is RECOMMENDED that: Server A Registrant should not expect the ENUM client to ignore NAPTRs with higher ORDER field values - the "winning" ones may have been discarded. Server A Registrant should not expect ENUM clients to conform to the ORDER and PREFERENCE/PRIORITY sort order he or she has specified for NAPTRs; end users may have their own preferences for ENUMservices. Client Each ENUM client may reorder the NAPTRs it receives only to match an explicit preference pre-specified by its end user. Client ENUM clients that offer a list of contacts to the end user for his or her choice may present all NAPTRs, not just the ones with the highest currently unprocessed ORDER field value. Server A Registrant should not assume which NAPTR choices will be presented "at once". The impact of this is that a Registrant should place into his or her zone only contacts that he or she is willing to support; even those with the "least preferred" ORDER and PREFERENCE/PRIORITY values may be selected by an end user. Finally, we have noticed a number of ENUM zones with NAPTRs that have Conroy & Fujiwara Expires December 27, 2006 [Page 12] Internet-Draft ENUM Experiences June 2006 identical PREFERENCE/PRIORITY field values and different ORDER values. This may be the result of an ENUM zone provisioning system "bug" or a misunderstanding over the uses of the two fields. To clarify, the ORDER field value is the major sort term, and the PREFERENCE/PRIORITY field value is the minor sort term. Thus one should expect to have a set of NAPTRs in a zone with identical ORDER field values and different PREFERENCE/PRIORITY field values. 4.2. NAPTRs with identical ORDER/PRIORITY values From experience, there are zones that hold discrete NAPTRs with identical ORDER and identical PREFERENCE/PRIORITY field values. This will lead to indeterminate client behaviour and so should not occur. However, in the spirit of being liberal in what is allowed: It is RECOMMENDED that: Client ENUM clients should accept all NAPTRs with identical ORDER and identical PREFERENCE/PRIORITY field values, and process them in the sequence in which they appear in the DNS response. (There is no benefit in further randomising the order in which these are processed, as intervening DNS Servers may do this already). Conversely, populating the records with these identical values is unwise, as it may lead to indeterminate client behaviour, and so it is RECOMMENDED that: Server When populating ENUM zones with NAPTRs, ENUM zone provisioning systems should not have more than one NAPTR with the same ORDER and the same PREFERENCE/PRIORITY field values in any given zone, as ENUM clients may reject the response, and the sequence in which these NAPTRs are delivered to the client may vary. 4.2.1. Compound NAPTRs and implicit ORDER/REFERENCE Values There is one special case in which one could derive a set of NAPTRs with identical ORDER and identical PREFERENCE/PRIORITY fields. This will not exist explicitly in the Resource Record Set ("RRSet") delivered to the client, but may occur whilst processing a "Compound" NAPTR, and is dealt with next. Conroy & Fujiwara Expires December 27, 2006 [Page 13] Internet-Draft ENUM Experiences June 2006 4.3. Compound NAPTR Processing With RFC 3761, it is possible to have more than one ENUMservice associated with a single NAPTR. Of course, the different ENUMservices share the same RegExp field and so generate the same URI. Such a "compound" NAPTR could well be used to indicate, for example, a mobile phone that supports both "voice:tel" and "sms:tel" ENUMservices. This compound NAPTR may be reconstructed into a set of NAPTRs each holding a single ENUMservice. However, in this case the members of this set all logically hold the same ORDER and PREFERENCE/PRIORITY field values. In this case, it is RECOMMENDED that: Client ENUM clients receiving compound NAPTRs (i.e. ones with more than one ENUMservice) should process these ENUMservices using a left-to-right sort ordering, so that the first ENUMservice to be processed will be the leftmost one, and the last will be the rightmost one. Server An ENUM zone provisioning system should assume that, if it generates compound NAPTRs, the ENUMservices will normally be processed in left to right order within such NAPTRs. As a final point on ENUM client processing of compound NAPTRs, it is quite possible that the client is incapable of processing one of the ENUMservices indicated. To clarify, it is RECOMMENDED that: Client When an ENUM client encounters a compound NAPTR and cannot process one of the ENUMservices within it, that ENUM client should ignore it and continue with the next ENUMservice within this NAPTR's Services field, discarding the NAPTR only if it cannot handle any of the ENUMservices contained. 4.4. Processing Order value across Zones Using a different ORDER field value in different zones is unimportant for most queries. However, DDDS includes a mechanism for continuing a search for NAPTRs in another zone by including a reference to that other zone in a "non-terminal" NAPTR. The treatment of non-terminal NAPTRs is covered in the next section, but if these are supported then it does have a bearing on the way that ORDER and PREFERENCE/ PRIORITY field values are processed. Conroy & Fujiwara Expires December 27, 2006 [Page 14] Internet-Draft ENUM Experiences June 2006 Two main questions remain from the specifications of DDDS and RFC 3671: o If there is a different (lower) order field value in a zone referred to by a non-terminal NAPTR, then does this mean that the ENUM client discards any remaining NAPTRs in the referring zone? o Conversely, if the zone referred to by a non-terminal NAPTR contains entries that have a higher ORDER field value, then does the ENUM client ignore those NAPTRs in the referenced zone? Whilst one interpretation of section 1.3 of RFC 3761 is that the answer to both questions is "yes", this is not the way that those examples of non-terminal NAPTRs that do exist (and those ENUM clients that support them) seem to be designed. Thus, to reflect the interpretation that is made by those systems that have implemented non-terminal NAPTRs, it is RECOMMENDED that: Server ENUM zone provisioning systems should assume that, once a non-terminal NAPTR has been selected for processing, the ORDER field value in a zone referred to by that non- terminal NAPTR will be considered only within the context of that referenced zone (i.e. the ORDER value will be used only to sort within the current zone, and will not be used in the processing of NAPTRs in any other zone). Client ENUM clients should consider the ORDER field value only when sorting NAPTRs within a single RRSet. The ORDER field value should not be taken into account when processing NAPTRs across a sequence of DNS queries created by traversal of non-terminal NAPTR references. Conroy & Fujiwara Expires December 27, 2006 [Page 15] Internet-Draft ENUM Experiences June 2006 5. Non-Terminal NAPTR Processing 5.1. Non-Terminal NAPTRs - necessity Consider an ENUM RRSet that contains a non-terminal NAPTR record. This non-terminal NAPTR "points to" another domain that has a set of NAPTRs. In effect, this is similar to the non-terminal NAPTR being replaced by the NAPTRs contained in the domain to which it points. It is possible to have a non-terminal NAPTR in a domain that is, itself, pointed to by another non-terminal NAPTR. Thus a set of domains forms a "chain", and the list of NAPTRs to be considered is the set of all NAPTRs contained in all of the domains in that chain. For an ENUM management system to support non-terminal NAPTRs, it is necessary for it to be able to analyse, validate and (where needed) correct not only the NAPTRs in its current ENUM domain but also those "pointed to" by non-terminal NAPTRs in other domains. If the domains pointed to have non-terminal NAPTRs of their own, the management system will have to check each of the referenced domains in turn, as their contents forms part of the result of a query on the "main" ENUM domain. The domain content in the referenced domains may well not be under the control of the ENUM management system, and so it may not be possible to correct any errors in those RRSets. This is both complex and prone to error in the management system design, and any reported errors in validation may well be non-intuitive for users. For an ENUM client, supporting non-terminal NAPTRs can also be difficult. Processing non-terminal NAPTRs causes a set of sequential DNS queries that can take an indeterminate time, and requires extra resources and complexity to handle fault conditions like non-terminal loops. The indeterminacy of response time makes ENUM supported Telephony Applications difficult (such as in an "ENUM-aware" PBX), whilst the added complexity and resources needed makes support problematic in embedded devices like "ENUM-aware" mobile phones. Given that, in principle, a non-terminal NAPTR can be replaced by the NAPTRs in the domain to which it points, support of non-terminal NAPTRs is not needed and non-terminal NAPTRs may not be useful. Furthermore, most existing ENUM clients do not support non-terminal NAPTRs and ignore them if received. To avoid interoperability problems, some kind of acceptable requirement is needed on non- terminal NAPTRs. Given the lack of current support and the issues raised, we propose that in general one should not use non-terminal NAPTRs in ENUM. Thus, it is RECOMMENDED that: Conroy & Fujiwara Expires December 27, 2006 [Page 16] Internet-Draft ENUM Experiences June 2006 Server ENUM zone provisioning systems should not generate non- terminal NAPTRs (i.e. NAPTRs with an empty Flags field) unless it is clear that all ENUM clients they are designed to support can process these. Client ENUM clients may discard non-terminal NAPTRs (i.e. they may only support ENUM NAPTRs with a Flags field value of "u"). 5.2. Non-Terminal NAPTRs - future implementation The following specific issues need to be considered if non-terminal NAPTRs are to be supported in the future. These issues are gleaned from experience, and indicate the kinds of conditions that should be considered before support for non-terminal NAPTRs is contemplated. Note that these issues are in addition to the point just mentioned on ENUM provisioning or management system complexity and the potential for that management system to have no control over the zone contents to which non-terminal NAPTRs in "its" managed zones refer. 5.2.1. Non-Terminal NAPTRs - general As mentioned earlier, a non-terminal NAPTR in one zone refers to the NAPTRs contained in another zone. The NAPTRs in the zone referred to by the non-terminal NAPTR may have a different ORDER value from that in the referring non-terminal NAPTR. See Section 4.4 for details. In addition, to Clarify, it is RECOMMENDED that: Client If all NAPTRs in a domain traversed as a result of a reference in a non-terminal NAPTR have been discarded, then the ENUM client should continue its processing with the next NAPTR in the "referring" RRSet (i.e. the one including the non-terminal NAPTR that caused the traversal). 5.2.2. Non-Terminal NAPTRs - loop detection and response Where a "chain" of non-terminal NAPTRs refers back to a domain already traversed in the current query, this implies a "non-terminal loop". To ensure consistent behaviour, it is RECOMMENDED that: Client ENUM clients should consider that processing a chain of more than 5 "non-terminal" NAPTRs in a single ENUM query indicates that a loop may have been detected, and act accordingly. Conroy & Fujiwara Expires December 27, 2006 [Page 17] Internet-Draft ENUM Experiences June 2006 Server When populating a set of domains with NAPTRs, ENUM zone provisioning systems should not configure non-terminal NAPTRs so that more than 5 such NAPTRs will be processed in an ENUM query. Client Where a domain is about to be entered as the result of a reference in a non-terminal NAPTR, and the ENUM client has detected a potential "non-terminal loop", then the client should discard the non-terminal NAPTR from its processing and continue with the next NAPTR in its list. It should not make the DNS query indicated by that non-terminal NAPTR. 5.3. Interpretation of RFC 3403 and RFC 3761 The set of specifications defining DDDS and its applications are complex and multi-layered. This reflects the flexibility that the system provides, but it does mean that some of the specifications need clarification as to their interpretation, particularly where non-terminal rules are concerned. 5.3.1. Flags field content with Non-Terminal NAPTRs RFC 3761, section 2.4.1 states that the only flag character valid for use with the "E2U" DDDS Application is 'u'. The flag 'u' is defined (in RFC 3404 [5], section 4.3) thus: 'The "u" flag means that the output of the Rule is a URI'. RFC 3761 section 2.4.1 also states that an empty Flags field indicates a non-terminal NAPTR. This is also the case for other DDDS Application specifications, such as that specified in RFC 3404. One could well argue that this is a feature potentially common to all DDDS Applications, and so should have been specified in RFC 3402 or RFC 3403. 5.3.2. Services field content with Non-Terminal NAPTRs Furthermore, RFC 3761 section 3.1.1 states that any ENUMservice Specification requires definition of the URI that is the expected output of this ENUMservice. This means that, at present, there is no way to specify an ENUMservice that is non-terminal. Such a non- terminal NAPTR has, by definition, no URI as its expected output, instead returning a key (DNS domain name) that is to be used in the "next round" of DDDS processing. This in turn means that there can be no valid (non-empty) Services Conroy & Fujiwara Expires December 27, 2006 [Page 18] Internet-Draft ENUM Experiences June 2006 field content for a NAPTR to be used with the "E2U" DDDS application. Section 2.4.2 of RFC 3761 specifies the syntax for this field content, and requires at least one element of type (i.e. at least one ENUMservice identifier). Given that there can be no definition of a non-terminal ENUMservice (and so no such Registered ENUMservice identifier), this syntax cannot be met with a non-terminal NAPTR. A reasonable interpretation of the specifications in their current state is that the Services field must also be empty; this appears to be the approach taken by those clients that do either process non- terminal NAPTRs or check the validity of the fields. To ensure consistent behaviour, it is RECOMMENDED that: Client ENUM clients should ignore any content of the Services field when encountering a non-terminal NAPTR with an empty Flags field. Server ENUM zone provisioning systems should ensure that the Services field of any non-terminal NAPTR (with an empty Flags field) is also empty. 5.3.3. Regular Expression and Replacement field content with non- terminal NAPTRs The descriptive text in section 4.1 of RFC 3403 is intended to explain how the fields are to be used in a NAPTR. However, the descriptions associated with the RegExp and Replacement elements have led to some confusion over which of these should be considered when dealing with non-terminal NAPTRs. RFC 3403 is specific; these two elements are mutually exclusive. This means that if the RegExp element is not empty then the Replacement element must be empty, and vice versa. However, is does not specify which is used with terminal and non-terminal rules. The descriptive text of section 4.1 of RFC 3403 for the NAPTR Replacement element shows that this element holds an uncompressed domain name. Thus it is clear that this element cannot be used to deliver the terminal string for any DDDS application that does not have a domain name as its intended terminal output. However, the first paragraph of descriptive text for the NAPTR RegExp element has led to some confusion. It appears that the RegExp element is to be used to find "the next domain name to lookup". This might be interpreted as meaning that a client program processing the DDDS application could need to examine each non-terminal NAPTR to decide whether the RegExp element or instead the Replacement element Conroy & Fujiwara Expires December 27, 2006 [Page 19] Internet-Draft ENUM Experiences June 2006 were to be used to construct the key (a domain name) to be used next in non-terminal rule processing. Given that a NAPTR holding a terminal rule (a "terminal NAPTR") must use the Substitution expression field to generate the expected output of that DDDS application, the RegExp element is also used in such rules. Indeed, unless that DDDS application has a domain name as its terminal output, the RegExp element is the only possibility. Thus from the descriptive text of this section, a Replacement element can be used only in NAPTRs holding a non-terminal rule (a "non- terminal NAPTR") unless that DDDS Application has a domain name as its terminal output, whilst the alternative RegExp element may be used either to generate a domain name as the next key to be used in the non-terminal case, or to generate the output of the DDDS application. Note that each DDDS Application is free to specify the set of flags to be used with that application. This includes specifying whether a particular flag is associated with a terminal or non-terminal rule, and also to specify the interpretation of an empty Flags field (i.e. whether this is to be interpreted as a terminal or non-terminal rule, and if it is terminal, then the expected output). ENUM (as specified in section 2.4.1 of RFC 3761) specifies only the 'u' flag, with an empty Flags field indicating a non-terminal NAPTR. The general case in which a client program must check which of the two elements to use in non-terminal NAPTR processing complicates implementation, and this interpretation has NOT been made in current ENUM examples "out in the wild". It would be useful to define exactly when a client program can expect to process the RegExp element and when to expect to process the Replacement element, if only to improve robustness. In keeping with current implementations, we suggest that a non- terminal NAPTR with an empty Flags field must be provisioned using the (non-empty) Replacement element to hold the domain name that forms the "next key" output from this non-terminal rule. Thus it is RECOMMENDED that: Client ENUM clients receiving a non-terminal NAPTR with an empty Flags field must treat the Replacement field as holding the domain name to be used in the next round of the ENUM query. An ENUM client must discard such a non-terminal NAPTR if the Replacement field is empty or does not contain a valid domain name. By definition, it follows that the RegExp field will be empty in such a non-terminal NAPTR, and Conroy & Fujiwara Expires December 27, 2006 [Page 20] Internet-Draft ENUM Experiences June 2006 should be ignored by ENUM clients Server with an empty Flags field into an ENUM zone must ensure that the "target" domain name is set into the Replacement field of this NAPTR. It must not use the RegExp field in such a non-terminal NAPTR. In the future, it would be possible to update RFC 3761 (sections 3.1.1 and 2.4.1) to add a new flag to indicate a non-terminal NAPTR, and to change the ENUMservice template to permit specification of an ENUMservice that operates with this new flag in non-terminal NAPTRs. In doing this, it would be possible to include a syntactically valid non-empty Services field in such non-terminal NAPTRs. To differentiate from the case of an empty Flags field, this new flag could also indicate that the RegExp field was to be non-empty, and to be processed - by implication, this would mean that the Replacement field would be empty. However, such a change would require an update to RFC 3761, and so will have to wait. Conroy & Fujiwara Expires December 27, 2006 [Page 21] Internet-Draft ENUM Experiences June 2006 6. General DNS Issues Whilst these issues covered in this section are not ENUM-specific, they do "bite" when developing ENUM-aware systems, and are important for ENUM deployment. This section makes strong recommendations over support for EDNS0 and TCP. These are crucial, as they will have a major impact on ENUM deployments, particularly when ENUM Clients are connected to the Internet over access networks that exhibit higher latencies (such as cellular mobile networks, where latencies can be of the order of hundreds of milliseconds or more). In addition, there are several elements of behaviour of the DNS that can complicate testing. As advice to developers, some of the more subtle issues are covered here, gleaned from the experience of those groups already developing and deploying ENUM systems. 6.1. DNS Specifications The DNS protocol is defined in RFC 1034 [8] and RFC 1035 [9], and is clarified in RFC 2181 [16], whilst Requirements for Internet Hosts are specified in RFC 1123 [17]. Security threats to DNS are covered in RFC 3833 [21], and DNSSEC (DNS Security) is specified in RFC 4033 [22], RFC 4034 [23], and in RFC 4035 [24], whilst EDNS0 (an extension mechanism for DNS) is specified in RFC 2671 [18]. Supporting UDP queries is mandatory, but support for TCP queries is recommended also, and is (in effect) required as RFC 1123 requires that a DNS client discard a truncated response sent in response to its initial (UDP) query and retry using another transport protocol. In effect, Authoritative Name Servers that do not answer TCP queries after returning truncated responses are not able to answer queries correctly. Amongst other things, the EDNS0 mechanism allows the querying client to indicate the size of UDP packet that it can process. RFC 1035 restricts the maximum size of a UDP DNS response to 512 bytes. This limit has proven to be too low for reasonably sized NAPTR Resource Record Sets. EDNS0 provides a mechanism for DNS clients and servers to use larger UDP payloads. Partially due to the larger response sizes involved and the need for extended flags, support for DNSSEC also requires the use of EDNS0. 6.2. ENUM needs EDNS0 support The performance of ENUM resolution is much reduced if a mechanism to handle larger responses efficiently is not available. ENUM RRSets often have more records than will fit into a basic "RFC 1035" UDP response. EDNS0 (as specified in RFC 2671 [18]) with a suitably Conroy & Fujiwara Expires December 27, 2006 [Page 22] Internet-Draft ENUM Experiences June 2006 large buffer size is the current official mechanism to support these larger messages with maximum efficiency. Lack of EDNS0 support and use of its size option by querying clients causes degradation of the system by requiring retransmission via TCP once a basic RFC 1035 UDP query results a "truncated" error response from the queried DNS Server. As ENUM RRSets typically contain more records than are returned for other queries, this is a much more frequent condition. Processing truncated responses adds latency to ENUM resolution and may consume excessive TCP resources on the DNS Server, such as data structures needed to maintain connection state. (see section 4.2.2 of [9] for the requirement for a DNS server to leave each TCP port used for queries open for several minutes, otherwise relying on the client to close the port explicitly when no longer needed). The impact of larger response messages where EDNS0 is not used can be significant. Instead of two messages (the query indicating EDNS0 support, and its response, using EDNS0), if the method described in RFC 1035 and RFC 1123 is used there will be many more - two messages for the query and the truncated UDP response, several messages to make the TCP connection, at least one for the query, and at least one for the response, with several more messages being needed to tear down the TCP connection. Especially if these messages are exchanged via an access network with significant latency for data transport, the total time needed to perform ENUM resolution can be excessive (of the order of seconds over a cellular access network), and much more than would be needed for the UDP query. In practice, not using EDNS0 where clients may be connected via these networks means that ENUM deployment will not meet acceptable performance targets for RRSets with larger sets of records. Note that this is not ENUM-specific but instead a standard feature of DNS - it just comes into play more often with ENUM and other queries that return large response messages. Thus to ensure efficient resolution of ENUM queries, it is RECOMMENDED that: Server All servers involved in ENUM resolution that carry messages including NAPTR RRsets must support EDNS0. Client ENUM Clients should use EDNS0 in their queries. The sole exception to this is when the Client knows that a server does not support EDNS0 (due to some misconfiguration). This adaptation is normal behaviour for a DNS Resolver that supports EDNS0 (see section 5 of [18]). Conroy & Fujiwara Expires December 27, 2006 [Page 23] Internet-Draft ENUM Experiences June 2006 6.3. ENUM EDNS0 message size support The next issue has no deterministic answer. Whilst it is impossible to be definitive over the size of messages exchanged during ENUM resolution, it is possible to state that the message sizes are considerably larger than has typically been the case for most DNS queries so far. Selecting appropriate sizes for EDNS0 support aims to ensure that the percentage of queries that still require a "fallback" to TCP is minimal for all realistically populated ENUM RRSets. In practice, zones "seen" so far have rarely required more than 2 kilobytes in response size. However, with the introduction of IPv6 and DNSSEC support, this is very likely to increase in the near future, so we recommend that ENUM Clients should indicate their support for a larger buffer size than might be necessary at this point. It should be noted that the penalty of choosing too low a size for EDNS0 support may be even more severe that the standard method described in RFC 1035 and RFC 1123. Thus it is good practice to select a larger size than is likely to be needed, to counteract that greater cost where fallbacks still occur. Sections 2.4 and particularly 2.5 of [24] explain the rationale for using the size option of EDNS0 for queries that return larger responses. In that document, section 3.1 describes expected server behaviour, section 4.1 describes expected Resolver behaviour, whilst section 3 summarises the proposed message sizes to be supported by Servers and Resolvers. These same size recommendations are repeated here, as it is felt that ENUM already has a similar issue with larger responses, and will certainly need the larger messages sizes with the introduction of IPv6 and DNSSEC support. It is RECOMMENDED that: Client An ENUM Client should be willing to handle DNS responses of up to 4000 bytes in length, and should indicate in its queries that it supports responses up to 4000 bytes in length using the EDNS0 technique specified in [18]. It is further RECOMMENDED that: Client An ENUM Client should consider an indicated supported value of 1220 bytes in its EDNS0 queries to be a bare minimum whilst avoiding fragmentation of response packets over most deployed networks. Conroy & Fujiwara Expires December 27, 2006 [Page 24] Internet-Draft ENUM Experiences June 2006 6.4. Intermediary Devices From our experiences a Name Server may support TCP queries, but there may well be an intervening packet filter that does not allow TCP traffic to pass correctly. It is unfortunately common for people managing a firewall to block traffic to or from the DNS TCP port without considering the impact. Thus if TCP queries do not seem to work, it is worthwhile considering this possibility; the Name Server may be operating correctly, but the TCP SYN or SYN-ACK packets may be blocked, effectively disabling the Server from contact with the World beyond the firewall. An incorrect assumption is made by some deployed packet filters that MAY affect transport of EDNS0 responses. It has been noticed that some older equipment may be configured by default to discard all UDP packets containing DNS messages if these are more than 512 bytes in size. Since the introduction of EDNS0 in 1999, such a configuration has been and is incorrect. Both of these behaviours can be very hard to debug. In particular, odd behaviour has been detected within several cellular networks that claim to provide Internet access; it appears that these networks include firewalls or application level proxies that do not handle larger DNS messages carried over UDP well. This is a basic error in network infrastructure configuration that may affect their customers, particularly when these customers attempt ENUM resolution. Experimental results have shown that such access network components typically seem to behave as "transparent" proxies for all traffic sent to the default DNS port (port 53), discarding larger packets regardless of content. Thus, although it should be obvious, we RECOMMEND that: MidBox Name Servers should support TCP queries, and must support EDNS0 if they are to host ENUM data. Thus intermediate systems such as firewalls should not be configured to filter traffic to or from a Name Server; notably, these should not block TCP transport for DNS queries, and must not simply block DNS messages of greater than 512 bytes in size without examining them for correct EDNS0 support. Note that this may require stateful packet inspection. Unfortunately, it is still necessary to note that: Client Developers of ENUM clients should be aware of these potential problems and clients must be able to adapt to such misconfigured systems, in accordance with RFC 2671 [18]. However, developers should also be aware that, in Conroy & Fujiwara Expires December 27, 2006 [Page 25] Internet-Draft ENUM Experiences June 2006 the case of malfunctioning intermediary systems, the proposal made there (to "probe" for DNS server capability prior to placing requests) may not succeed in exposing an infrastructure that will disable normally functioning servers. Small "probe" messages may pass, indicating EDNS0 support, whilst longer messages in subsequent queries do not. 6.5. Times To Live and NAPTRs Section 5.2 of [16] clarifies the normal operation of DNS where more than one Resource Record (RR) is returned in a response. It defines the requirement that Time To Live (TTL) values must be the same in those circumstances, and the treatment by Resolvers of response messages containing sets of Resource Records. As a query on an ENUM zone for NAPTRs will typically return a set of resource records consisting of more than one record, this has an impact on ENUM processing, and it is worth exploring it in this document. It is important that all NAPTRs in a zone have Time To Live fields set to the same value, or else "strange things" will happen with Recursive Resolvers that are hard to debug. A query for type NAPTR will return the available collection of such RRs. From the perspective of the Recursive Resolver, if records in this set have different TTL, then the zone will be re-queried only once ALL queried Records have expired. Given this behaviour by DNS Resolvers, it is unwise to have different TTL in the various NAPTRs in a zone. Some of these resource records will seem to "disappear" over time from the list of available records returned and that will lead to variable and often unexpected behaviour - if nothing else, it may well confuse the end user as the list of options he or she is offered grows and shrinks. This is true for RR type-specific queries, but it is also true for general queries within a zone. For example, a DNS query with a Query Type value of 255 asks for "ANY Resource Records for that domain name that you you have". Thus if a Recursive Resolver holds RRs for a zone in its cache, and a cached RR from that zone has a long TTL, then it will be returned even after others have expired - no onward query will be made unless there are no RRs at all available at the Recursive Resolver for the queried zone. In this case, however, the TTL of ALL the records that may well exist in a queried zone have to be taken into account, rather than just those of the NAPTR records themselves. This may well not be the case, so this kind of query should be avoided. In summary, in keeping with the standards: Conroy & Fujiwara Expires December 27, 2006 [Page 26] Internet-Draft ENUM Experiences June 2006 Server ENUM provisioning systems must follow the guidance of section 5.2 of RFC 2181 [16]. All TTL values must be identical for Resource Records of a given class and type. Client ENUM clients must follow the guidance of section 5.2 of RFC 2181 [16]. Client ENUM clients should not issue queries with Query Type = 255 (commonly known as an 'ANY' query), as the results may well be unexpected. Conroy & Fujiwara Expires December 27, 2006 [Page 27] Internet-Draft ENUM Experiences June 2006 7. Backwards Compatibility 7.1. Services field syntax RFC 3761 is the current standard for the syntax for NAPTRs supporting the ENUM DDDS application. This obsoletes the original specification that was given in RFC 2916. There has been a change to the syntax of the Services field of the NAPTR that reflects a refinement of the concept of ENUM processing. As defined in RFC 3403, there is now a single identifier that indicates the DDDS Application. In the obsolete specification (RFC 2915), there were zero or more "Resolution Service" identifiers (the equivalent of the DDDS Application). The same identifier string is defined in both RFC 3761 and in the old RFC 2916 specifications for the DDDS identifier or the Resolution Service; "E2U". Also, RFC 3761 defines at least one but potentially several ENUMservice sub-fields; in the obsolete specification, only one "protocol" sub-field was allowed. In many ways, the most important change for implementations is that the order of the sub-fields has been reversed. RFC 3761 specifies that the DDDS Application identifier is the leftmost sub-field, followed by one or more ENUMservice sub-fields, each separated by the '+' character delimiter. RFC 2916 specified that the protocol sub- field was the leftmost, followed by the '+' delimiter, in turn followed by the "E2U" resolution service tag. RFC 2915 and RFC 2916 have been obsoleted by RFC 3401 - RFC 3404 and by RFC 3761. Thus it is RECOMMENDED that: Server ENUM zone provisioning systems must not generate NAPTRs according to the syntax defined in RFC 2916. All zones must hold ENUM NAPTRs according to RFC 3761 (and ENUMservice specifications according to the framework specified there). However, RFC 3824 [19] suggests that ENUM clients should be prepared to accept NAPTRs with the obsolete syntax. Thus, an ENUM client implementation may have to deal with both forms. It is RECOMMENDED that: Client ENUM clients must support ENUM NAPTRs according to RFC 3761 syntax. ENUM clients should also support ENUM NAPTRs according to the obsolete syntax of RFC 2916; there are still zones that hold "old" syntax NAPTRs. Conroy & Fujiwara Expires December 27, 2006 [Page 28] Internet-Draft ENUM Experiences June 2006 This need not be difficult. For example, an implementation could process the Services field into a set of tokens, and expect exactly one of these tokens to be "E2U". In this way, the ENUM client might be designed to handle both the old and the current forms without added complexity. There is one subtle implication of this scheme. It is RECOMMENDED that: Spec Registrations for an ENUMservice with the type string of "E2U" and an empty sub-type string must not be accepted. Conroy & Fujiwara Expires December 27, 2006 [Page 29] Internet-Draft ENUM Experiences June 2006 8. Security Considerations This document does not specify any standard. It does however make some recommendations, and so the implications of following those suggestions have to be considered. In addition to these issues, those in the basic use of ENUM (and specified in the normative documents for this protocol) should be considered as well; this document does not negate those in any way. The clarifications throughout this document are intended only as that; clarifications of text in the normative documents. They do not appear to have any security implications above those mentioned in the normative documents. The suggestions in Section 3, Section 4, and Section 7 do not appear to have any security considerations (either positive or negative). The suggestions in Section 5.2.2 are a valid approach to a known security threat. It does not open an advantage to an attacker in causing excess processing or memory usage in the client. It does, however, mean that an ENUM client will traverse a "tight loop" of non-terminal NAPTRs in two domains 5 times before the client detects this as a loop; this does introduce slightly higher processing load than would be provided using other methods, but avoids the risks they incur. Conroy & Fujiwara Expires December 27, 2006 [Page 30] Internet-Draft ENUM Experiences June 2006 9. IANA Considerations This document is only advisory, and does not include any IANA considerations other than the proposals labelled as "Spec". These are the recommendation (in Section 3.1) that ENUMservice Registrations should at least indicate if characters outside of the US-ASCII equivalent range are permitted, and the suggestion (at the end of Section 7.1) that no-one should specify an ENUMservice with the identifying tag "E2U". Conroy & Fujiwara Expires December 27, 2006 [Page 31] Internet-Draft ENUM Experiences June 2006 10. Acknowledgements We would like to thank the various development teams who implemented ENUM (both creation systems and clients) and who read the normative documents differently - without these differences it would have been harder for us all to develop robust clients and suitably conservative management systems. We would also thank those who allowed us to check their implementations to explore behaviour; their trust and help were much appreciated. In particular, thanks to Richard Stastny for his hard work on a similar task TS 102 172 [25] under the aegis of ETSI, and for supporting some of the ENUM implementations that exist today. Finally, thanks for the dedication of Michael Mealling in giving us such detailed DDDS specifications, without which the ENUM development effort would have had a less rigourous framework on which to build. This document reflects how complex a system it is: Without the intricacy of RFC 3401 - RFC 3404 and the work that went into them, it could have been quite different. Conroy & Fujiwara Expires December 27, 2006 [Page 32] Internet-Draft ENUM Experiences June 2006 11. References 11.1. Normative References [1] Faltstrom, P. and M. Mealling, "The E.164 to Uniform Resource Identifiers (URI) Dynamic Delegation Discovery System (DDDS) Application (ENUM)", RFC 3761, April 2004. [2] Mealling, M., "Dynamic Delegation Discovery System (DDDS) Part Three: The Domain Name System (DNS) Database", RFC 3403, October 2002. [3] Mealling, M., "Dynamic Delegation Discovery System (DDDS) Part One: The Comprehensive DDDS", RFC 3401, October 2002. [4] Mealling, M., "Dynamic Delegation Discovery System (DDDS) Part Two: The Algorithm", RFC 3402, October 2002. [5] Mealling, M., "Dynamic Delegation Discovery System (DDDS) Part Four: The Uniform Resource Identifiers (URI)", RFC 3404, October 2002. [6] Mealling, M., "Dynamic Delegation Discovery System (DDDS) Part Five: URI.ARPA Assignment Procedures", RFC 3405, October 2002. [7] Yergeau, F., "UTF-8, a transformation format of ISO 10646", STD 63, RFC 3629, November 2003. [8] Mockapetris, P., "DOMAIN NAMES - CONCEPTS AND FACILITIES", RFC 1034, November 1987. [9] Mockapetris, P., "Domain names - implementation and specification", STD 13, RFC 1035, November 1987. [10] Costello, A., "Punycode: A Bootstring encoding of Unicode for Internationalized Domain Names in Applications (IDNA)", RFC 3492, March 2003. [11] Institute of Electrical and Electronics Engineers, "Information Technology - Portable Operating System Interface (POSIX) - Part 2: Shell and Utilities (Vol. 1)", IEEE Standard 1003.2, January 1993. [12] Schulzrinne, H., "The tel URI for Telephone Numbers", RFC 3966, December 2004. [13] ITU-T, "The International Public Telecommunication Number Plan", Recommendation E.164, May 1997. Conroy & Fujiwara Expires December 27, 2006 [Page 33] Internet-Draft ENUM Experiences June 2006 [14] Berners-Lee, T., Fielding, R., and L. Masinter, "Uniform Resource Identifier (URI): Generic Syntax", RFC 3986, January 2005. [15] Duerst, M. and M. Suignard, "Internationalized Resource Identifiers (IRIs)", RFC 3987, January 2005. [16] Elz, R. and R. Bush, "Clarifications to the DNS Specification", RFC 2181, July 1997. [17] Braden, R., "Requirements for Internet Hosts -- Application and Support", RFC 1123, October 1989. [18] Vixie, P., "Extension Mechanisms for DNS (EDNS0)", RFC 2671, August 1999. [19] Peterson, J., Liu, H., Yu, J., and B. Campbell, "Using E.164 numbers with the Session Initiation Protocol (SIP)", RFC 3824, June 2004. 11.2. Informative References [20] American National Standards Institute, "Coded Character Set -- 7-bit American Standard Code for Information Interchange", ANSI X3.4, 1986. [21] Atkins, D. and R. Austein, "Threat Analysis of the Domain Name System (DNS)", RFC 3833, August 2004. [22] Arends, R., Austein, R., Larson, M., Massey, D., and S. Rose, "DNS Security Introduction and Requirements", RFC 4033, March 2005. [23] Arends, R., Austein, R., Larson, M., Massey, D., and S. Rose, "Resource Records for the DNS Security Extensions", RFC 4034, March 2005. [24] Arends, R., Austein, R., Larson, M., Massey, D., and S. Rose, "Protocol Modifications for the DNS Security Extensions", RFC 4035, March 2005. [25] ETSI, "Minimum Requirements for Interoperability of European ENUM Implementations", ETSI TS 102 172, October 2004. Conroy & Fujiwara Expires December 27, 2006 [Page 34] Internet-Draft ENUM Experiences June 2006 Authors' Addresses Lawrence Conroy Roke Manor Research Roke Manor Old Salisbury Lane Romsey United Kingdom Phone: +44-1794-833666 Email: lconroy@insensate.co.uk URI: http://www.sienum.co.uk Kazunori Fujiwara Japan Registry Service Co., Ltd. Chiyoda First Bldg. East 13F 3-8-1 Nishi-Kanda Chiyoda-ku Tokyo 101-0165 JAPAN Email: fujiwara@jprs.co.jp URI: http://jprs.jp/en/ Conroy & Fujiwara Expires December 27, 2006 [Page 35] Internet-Draft ENUM Experiences June 2006 Intellectual Property Statement The IETF takes no position regarding the validity or scope of any Intellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; nor does it represent that it has made any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can be found in BCP 78 and BCP 79. Copies of IPR disclosures made to the IETF Secretariat and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of this specification can be obtained from the IETF on-line IPR repository at http://www.ietf.org/ipr. The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights that may cover technology that may be required to implement this standard. Please address the information to the IETF at ietf-ipr@ietf.org. Disclaimer of Validity This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Copyright Statement Copyright (C) The Internet Society (2006). This document is subject to the rights, licenses and restrictions contained in BCP 78, and except as set forth therein, the authors retain all their rights. Acknowledgment Funding for the RFC Editor function is currently provided by the Internet Society. Conroy & Fujiwara Expires December 27, 2006 [Page 36]