Network Working Group L. Conroy

Network Working Group L. Conroy
Request for Comments: 5483 RMRL
Category: Informational K. Fujiwara
JPRS
March 2009

ENUM Implementation Issues and Experiences

Status of This Memo

This memo provides information for the Internet community. It does
not specify an Internet standard of any kind. Distribution of this
memo is unlimited.

Copyright Notice

Copyright (c) 2009 IETF Trust and the persons identified as the
document authors. All rights reserved.

This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents in effect on the date of
publication of this document (http://trustee.ietf.org/license-info).
Please review these documents carefully, as they describe your rights
and restrictions with respect to this document.

Abstract

This document captures experiences in implementing systems based on
the ENUM protocol and experiences of ENUM data that have been created
by others. As such, it clarifies the ENUM and Dynamic Delegation
Discovery System standards. Its aim is to help others by reporting
both what is "out there" and potential pitfalls in interpreting the
set of documents that specify the ENUM protocol. It does not revise
the standards but is intended to provide technical input to future
revisions of those documents.

Conroy & Fujiwara Informational [Page 1]

RFC 5483 ENUM Experiences March 2009

Table of Contents

1. Introduction ....................................................3
1.1. Document Goal ..............................................3
1.2. Terminology ................................................3
2. Character Sets and ENUM .........................................4
2.1. Character Sets - Non-ASCII Considered Harmful ..............4
2.1.1. Non-ASCII in the Regular Expression Field ...........5
2.1.2. Non-ASCII Support - Conclusions .....................6
2.2. Case Sensitivity ...........................................7
2.3. Regexp Field Delimiter .....................................7
2.4. Regexp Meta-Character Issue ................................8
3. Unsupported NAPTRs ..............................................8
3.1. Non-Compliant Client Behaviour .............................9
4. ENUM NAPTR Processing ..........................................10
4.1. Common Non-Compliant Client Behaviour .....................11
4.1.1. Example ............................................11
4.2. Order/Priority Values - Processing Sequence ...............12
4.3. Use of Order and Preference Fields ........................13
4.4. NAPTRs with Identical ORDER/PRIORITY Values ...............14
4.4.1. Compound NAPTRs and Implicit
ORDER/REFERENCE Values .............................14
4.5. Processing Order Value across Domains .....................15
5. Non-Terminal NAPTR Processing ..................................16
5.1. Non-Terminal NAPTRs - Necessity ...........................16
5.2. Non-Terminal NAPTRs - Considerations ......................17
5.2.1. Non-Terminal NAPTRs - General ......................17
5.2.2. Non-Terminal NAPTRs - Loop Detection and Response ..17
5.2.3. Field Content in Non-Terminal NAPTRs ...............17
6. Backwards Compatibility ........................................20
6.1. Services Field Syntax .....................................20
7. Collected Implications for ENUM Provisioning ...................21
8. Collected Implications for ENUM Clients ........................23
8.1. Non-Terminal NAPTR Processing .............................25
9. Security Considerations ........................................26
10. Acknowledgements ..............................................27
11. References ....................................................27
11.1. Normative References .....................................27
11.2. Informative References ...................................29

Conroy & Fujiwara Informational [Page 2]

RFC 5483 ENUM Experiences March 2009

1. Introduction

1.1. Document Goal

The goal of this document is to clarify the ENUM and Dynamic
Delegation Discovery System (DDDS) standards. It does not itself
revise ENUM or DDDS standards but is intended to provide technical
input to future revisions of those documents. It also serves to
advise implementers on the pitfalls that they may find. It
highlights areas where ENUM implementations have differed over
interpretation of the standards documents or have outright failed to
implement some features as specified.

As well as providing clarifications to standards text, this document
also mentions potential choices that can be made, in an attempt to
help foster interworking between components that use this protocol.
The reader is reminded that others may make different choices.

The core specifications for the E.164 Number Mapping (ENUM) protocol
[RFC3761] and the Dynamic Delegation Discovery System (DDDS)
[RFC3403] [RFC3401] [RFC3402] [RFC3404] [RFC3405] are defined
elsewhere. 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.

The Domain Name System (DNS) is ENUM's database. ENUM uses the NAPTR
(Naming Authority Pointer) resource record type to store its DDDS
rules into DNS domains. ENUM relies on DNS services. Thus, it is
also important for ENUM implementers to carry out a thorough analysis
of all of the existing DNS standard documents to understand what
services are provided to ENUM and what load ENUM provisioning and
queries will place on the DNS.

A great deal of the rationale for making the choices listed in this
document is available to those who explore the standards. The trick
of course is in understanding those standards and the subtle
implications that are involved in some of their features. In almost
all cases, the choices presented here are merely selections from
values that are permissible within the standards.

1.2. Terminology

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in [RFC2119].

Conroy & Fujiwara Informational [Page 3]

RFC 5483 ENUM Experiences March 2009

2. Character Sets and ENUM

2.1. Character Sets - Non-ASCII Considered Harmful

[RFC3403] and [RFC3761] specify respectively that NAPTR resource
records and ENUM support Unicode using the UTF-8 encoding defined in
[RFC3629]. This raises an issue when 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 [ASCII] 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. Specifically, if NAPTR content
is shown to an end user so that he or she may choose, it is
imperative that the content is human-readable. Thus, it is unwise to
use non-printable characters even if they lie 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
because string processing should not be applied to them. The string-
based fields are the Flags, Services, and Regexp fields. The
Replacement field holds an uncompressed domain name, encoded
according to the standard DNS mechanism [RFC1034][RFC1035]. The
Internationalised Domain Name (IDN) can be supported (as specified in
[RFC3490], [RFC3491], and [RFC3492]). Any such IDN MUST be further
encoded using Punycode [RFC3492]. As the Replacement field 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

Conroy & Fujiwara Informational [Page 4]

RFC 5483 ENUM Experiences March 2009

character as specified in [RFC3761]. These characters all fall into
the printable US-ASCII equivalent range, so multi-byte characters
cannot occur.

The Services field includes the DDDS Application identifier ("E2U")
used for ENUM, a set of Enumservice identifiers, any of which may
embed the ':' separator character, together with the '+' character
used to separate Enumservices from one another and from this DDDS
Application identifier. In Section 2.4.2 of [RFC3761], Enumservice
identifier tokens are specified as 1*32 ALPHA/DIGIT, so there is no
possibility of non-ASCII characters in the Services field.

2.1.1. Non-ASCII in the Regular Expression Field

The Regexp field is more complex. It forms a sed-like substitution
expression, defined in [RFC3402], and consists of two sub-fields:

o a POSIX Extended Regular Expression (ERE) sub-field
[IEEE.1003-2.1992]

o a replacement (Repl) sub-field [RFC3402].

Additionally, [RFC3402] 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 Section 2.2.

The ERE sub-field matches against the "Application Unique String";
for ENUM, this is defined in [RFC3761] 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
[RFC3966], but with visual-separators removed. In short, it is a
telephone number (see [E.164]) 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.

Strictly, the ERE might include other characters. The ERE could
include choice elements matching against different items, some of
which might not be an ENUM Application Unique String. Those
alternative matching elements might conceivably include non-ASCII
characters. As an operational issue, it is not reasonable to include
such constructs, as ENUM NAPTRs match against telephone numbers.

In the normal situation in which E2U NAPTRs are provisioned in ENUM
domains, there will be no multi-byte characters within this sub-
field, as the ERE will be intended to match against telephone
numbers. ENUM clients must be able to handle NAPTRs that do contain
such multi-byte characters (as the standard does not preclude them),
but there is no operational reason for these ever being provisioned

Conroy & Fujiwara Informational [Page 5]

RFC 5483 ENUM Experiences March 2009

in ENUM domains. If NAPTRs provisioned in ENUM domains are
encountered containing such multi-byte characters, these could
reasonably be discarded.

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 [RFC3403]. 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 [RFC3986] 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
character will be pre-processed; they will not occur in the explicit
text used to construct a URI within the Repl sub-field.

2.1.1.1. Impact of Future Support for IRIs

As currently specified, ENUM only permits URIs to be generated in the
Regexp field. However, even if this were to be extended in future
revisions of the ENUM specification to allow the use of
Internationalised Resource Identifiers (IRIs), defined in [RFC3987],
further support for non-ASCII characters may be avoided. IRIs are
defined as extending the syntax of URIs, and RFC 3987 specifies a
mapping from IRIs to URIs. IRI syntax allows characters with multi-
byte UTF-8 encoding.

Given that this is the only place within an ENUM NAPTR where such
multi-byte encodings might reasonably be found, a simple solution is
to use the mapping method specified in Section 3.1 of [RFC3987] 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 [RFC3987] 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.

2.1.2. Non-ASCII Support - Conclusions

From the analysis just given, the only place within an ENUM NAPTR
where non-ASCII characters might be found is the Regexp field. It is
possible to remove any requirement to process characters outside the

Conroy & Fujiwara Informational [Page 6]

RFC 5483 ENUM Experiences March 2009

US-ASCII equivalent range by adding very few operational
restrictions. There is no obvious benefit in providing characters
outside this range. Handling multi-byte characters complicates
development and operation of client programs, and many existing
programs do not include such support.

As the gain from permitting characters outside the US-ASCII
equivalent range is unclear, and the costs of multi-byte character
processing are very clear, ENUM NAPTRs SHOULD NOT include characters
outside the printable US-ASCII equivalent range.

2.2. 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. This flag is redundant for E2U NAPTRs, as it does not
act on the Repl sub-field contents.

Thus, the case-sensitivity flag is inappropriate for ENUM, and SHOULD
NOT be provisioned into E2U NAPTRs.

2.3. Regexp Field Delimiter

It is not possible to select a delimiter character that cannot appear
in one of the sub-fields. The '!' character is used as a delimiter
in all of the examples in [RFC3403] and in [RFC3761]. It is the only
character seen in existing zones, and a number of different client
implementations are still "hardwired" to expect this character as a
delimiter.

The '!' character will not normally 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). If it is present in the Regexp field, then
that instance MUST be escaped using the standard technique proposed
in Section 3.2 of [RFC3402]: a backslash character (U+005C) should be
inserted before it in the string. Otherwise, a client may attempt to
process this as a standard delimiter and interpret the Regexp field
contents differently from the system that provisioned it.

Conroy & Fujiwara Informational [Page 7]

RFC 5483 ENUM Experiences March 2009

2.4. 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 (to produce the sub-string U+005C U+002B), as '+'
is a meta-character in POSIX Extended Regular Expression syntax.

Not escaping the '+' character produces an invalid ERE, but is a
common mistake. Even standards have given incorrect examples; the
obsolete [RFC2916] (Section 3.4.3, example 3) has this problem.

For example, the following NAPTR example is incorrect:
* IN NAPTR 100 10 "u" "E2U+sip" "!^+4655(.*)$!sip:\\[email protected]!" .

A correct way to write this example is:
* IN NAPTR 100 10 "u"
"E2U+sip" "!^\\+4655(.*)$!sip:\\[email protected]!" .

Note that when a NAPTR resource record is shown in DNS master file
syntax (as in this example above), the backslash itself must be
escaped using a second backslash. The DNS on-the-wire packet will
have only a single backslash.

3. Unsupported NAPTRs

An ENUM client MAY discard a NAPTR received in response to an ENUM
query because:

o the NAPTR is syntactically or semantically incorrect,

o the NAPTR has a different (non-empty) DDDS Application identifier
from the 'E2U' used in ENUM,

o the NAPTR's ERE does not match the Application Unique String for
this ENUM query,

o the ENUM client does not recognise any Enumservice held in this
NAPTR, or

o this NAPTR (only) contains an Enumservice that is unsupported.

These conditions SHOULD NOT cause the whole ENUM query to terminate,
and processing SHOULD continue with the next NAPTR in the returned
Resource Record Set (RRSet).

Conroy & Fujiwara Informational [Page 8]

RFC 5483 ENUM Experiences March 2009

When an ENUM client encounters a compound NAPTR (i.e., one containing
more than one Enumservice -- see also Section 4.4.1) and cannot
process or cannot recognise one of the Enumservices within it, that
ENUM client SHOULD ignore this Enumservice 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. These
conditions SHOULD NOT be considered errors.

ENUM uses regular-expression processing when generating URIs from the
Regexp field of "terminal" NAPTRs. Just as with all uses of regular
expressions, there is a potential for buffer overrun when generating
this output. There may be repeated back-reference patterns in a
NAPTR's Repl sub-field, and the output these generate may consume a
considerable amount of buffer space.

Even if an ENUM client would normally encounter only NAPTRs with
short URIs, it may also receive NAPTRs with repeated back-reference
patterns in their Repl sub-fields that could generate strings longer
than the client's buffer. Such NAPTRs may have been misconfigured
accidentally or by design. The client MUST NOT fail in this case.
It SHOULD NOT discard the entire ENUM query, but instead just discard
the NAPTR that would otherwise have caused this overrun.

If a problem is detected when processing an ENUM query across
multiple domains (by following non-terminal NAPTR references), then
the ENUM query SHOULD NOT be abandoned, but instead processing SHOULD
continue at the next NAPTR after the non-terminal NAPTR that referred
to the domain in which the problem would have occurred. See
Section 5.2.2 for more details.

3.1. Non-Compliant Client Behaviour

Through monitoring current ENUM clients, a number of non-compliant
behaviours have been detected. These behaviours are incorrect, but
may be encountered in still-operational client implementations.

ENUM clients have been known to discard NAPTRs in which the Services
field holds more than one Enumservice.

ENUM clients have also been known to discard NAPTRs with a "non-
greedy" ERE sub-field expression (i.e., EREs that are dissimilar to
"^.*$").

ENUM clients have been known to discard NAPTRs that do not use '!' as
their Regexp delimiter character.

ENUM clients have been known to discard NAPTRs in which the delimiter
is NOT the last character in the Regexp field.

Conroy & Fujiwara Informational [Page 9]

RFC 5483 ENUM Experiences March 2009

ENUM clients have been known to discard NAPTRs with an empty Flags
field (i.e., "non-terminal" NAPTRs).

ENUM clients have been known to ignore the ORDER field value
entirely, sorting the NAPTRs in an RRSet based solely on the
PREFERENCE/PRIORITY field values.

Finally, many ENUM clients have been known to discard a NAPTR where
they have local knowledge that the URI that would be generated by
processing the NAPTR is unusable. This behaviour is, strictly
speaking, non-compliant, but might be considered reasonable (see
Section 4.1).

4. ENUM NAPTR Processing

ENUM is a DDDS Application, and the way in which NAPTRs in an RRSet
are processed reflects this. The details are described in Section
3.3 of [RFC3402]. The client is expected to sort the records it
receives into a sequence and then process those records in that
sequence. The sequence reflects the ORDER and PREFERENCE/PRIORITY
field values in each of the NAPTRs.

The ORDER field value is the major, or most significant, sort term
and the PREFERENCE/PRIORITY field value is the minor, or least
significant, sort term. The combination of ORDER and PREFERENCE/
PRIORITY field values indicates the sequence chosen by the publisher
of this data, and NAPTRs will be considered in this sequence.

Once the NAPTRs are sorted into sequence, further processing is done
to determine if each of the NAPTRs is appropriate for this ENUM
evaluation. This involves looking at the Flags field. If the Flags
field is empty, this is a "non-terminal" NAPTR and is processed as
described in Section 5.

If the "u" Flag is present (and so the NAPTR is a "terminal" rule
that generates a URI), the Services field is checked to ensure that
this NAPTR is intended for ENUM (i.e., that this NAPTR includes the
"E2U" DDDS Application identifier in the Services field). The ERE in
the Regexp field is checked and must match the Application Unique
String (AUS) for this ENUM evaluation (the queried telephone number).
Unless each of these checks succeeds, the NAPTR is discarded and the
next in sequence is processed.

During this processing, clients will also consider the Enumservices
within the Services field. Enumservices indicate the kind of
interaction that can be achieved through use of the URI this NAPTR
generates. If there is local knowledge that a NAPTR includes only an
Enumservice that is either not supported or not recognised, then this

Conroy & Fujiwara Informational [Page 10]

RFC 5483 ENUM Experiences March 2009

NAPTR can be discarded and the next in sequence will be processed.
Thus, for a system that has support only for SIP interactions, if it
receives an RRSet in which the "best" NAPTR indicates the H323
Enumservice, then that client could reasonably discard that NAPTR and
go on to the next in sequence.

4.1. Common Non-Compliant ENUM Processing

The processing of ORDER and PREFERENCE/PRIORITY fields has been a
significant source of confusion, and many ENUM clients do not
implement the processing exactly as specified.

In particular, many ENUM clients use local prior knowledge about URIs
during ENUM processing. If a client has prior knowledge that a
particular URI will not result in an acceptable outcome, it might
discard that NAPTR and consider the next one in the sequence.
Examples of such local prior knowledge include: the URI does not
resolve, authentication has been recently rejected, or user policies
mark a particular URI as unacceptable (the URI could be a "premium
rate" telephone number that would be charged at an unacceptable
rate).

Strictly speaking, this behaviour is non-compliant if the next NAPTR
record has a different ORDER value. The ENUM algorithm (Section 3.3
of [RFC3402] and Section 4.1 of [RFC3403]) states that once a match
has been found for the Application Unique String (AUS), and the
service description satisfies the client's requirements, NAPTR
records with larger ORDER values must not be considered (but other
NAPTR records with the same ORDER value can still be considered).

However, embedding local knowledge about the URI within the ENUM
evaluation process is almost universal in systems employing ENUM.
Also, since the difference between ORDER and PRIORITY/PREFERENCE has
been unclear, NAPTR records have been provisioned in ways that would
make strictly compliant systems unusable in practice. Given that
such systems are intended to provide communications, this non-
compliant, "embedded decision" behaviour is understandable.

It is proposed that when the ENUM specification is updated,
processing of ORDER and PRIORITY/PREFERENCE should be updated based
on implementation and deployment experiences described in this
document.

4.1.1. Example

The example in this section is intended to further understanding
about the difference between what [RFC3402] and [RFC3403] specify and
what existing ENUM clients do.

Conroy & Fujiwara Informational [Page 11]

RFC 5483 ENUM Experiences March 2009

WARNING: The NAPTR records shown in this section are intended to
illustrate somewhat unclear corner cases, and are not intended as
good examples of how to do ENUM provisioning.

Consider the following RRset, which maps numbers in the UK drama
range to one server, and all other numbers to a second server:
* 3600 IN NAPTR 1 1 "u" "e2u+sip"
"!^(\\+441632960.*)$!sips:\\[email protected]!" .
* 3600 IN NAPTR 2 1 "u" "e2u+sip"
"!^(.*)$!sip:\\[email protected]!" .

According to the processing specified in [RFC3402] and [RFC3403], the
ENUM client is never intended to consider the second rule for e.g.,
AUS "+441632960123", even if it does not support "sips" URIs, or the
atlanta.example.com server cannot be reached, or the user indicates
he or she doesn't wish to contact atlanta.example.com. However,
existing ENUM implementations are known to do this, and as described
above, it can be useful if the alternative is failing to communicate
at all.

To prevent a client from considering the second rule for the UK drama
range, the example could be rewritten to have more predictable
behaviour as follows:
* 3600 IN NAPTR 1 1 "u" "e2u+sip"
"!^(\\+441632960.*)$!sips:\\[email protected]!" .
* 3600 IN NAPTR 2 1 "u" "e2u+sip"
"!^(\\+[^4].*|\\+4[^4].*|\\+44[^1].*|\\+441[^6].*|\\+4416[^3].*|
\\+44163[^2].*|\\+441632[^9].*|\\+4416329[^6].*|
\\+44163296[^0].*)$!sip:\\[email protected]!" .

4.2. Order/Priority Values - Processing Sequence

[RFC3761] and [RFC3403] 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 worse
ORDER values must only be dealt with once the current ones with a
better ORDER value have been processed.

However, as described in the introduction to this section, this
stated behaviour is a simplification. Once sorted into a sequence
reflecting ORDER and PREFERENCE/PRIORITY values, other fields are
also considered during evaluation of retrieved NAPTRs; local
knowledge may play a factor in the decision process, once a NAPTR has
reached that point in the sequence at which it is considered.

Conroy & Fujiwara Informational [Page 12]

RFC 5483 ENUM Experiences March 2009

ENUM clients may also include the end user "in the decision loop",
offering the end user the choice from a list of possible NAPTRs.
Conceptually this choice is embedded within step 4 of the DDDS
algorithm (as described in Section 3.3 of [RFC3402]). Given that the
ORDER field value is the major sort term, one would expect a
conforming ENUM client to present only those NAPTRs with the
currently "best" ORDER field value as choices. When/if all the
presented options 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
at once, in the sequence defined by the ORDER and PREFERENCE/PRIORITY
fields.

In summary, ENUM clients will take into account the Services field
value, the Flags field, and the Regexp ERE sub-field, along with the
ORDER and PREFERENCE/PRIORITY field values, and may consider local
policies or available local knowledge.

The Registrant and the ENUM zone provisioning system he or she uses
must be aware of this and SHOULD NOT rely on ENUM clients solely
taking account of the value of the ORDER and the PREFERENCE/PRIORITY
fields alone.

Specifically, it is unsafe to assume that an ENUM client will not
consider another NAPTR if there is one with a better ORDER value.
The instructions in Section 4.1 and Section 8 of [RFC3403] may or may
not be followed strictly by different ENUM clients for perfectly
justifiable reasons.

Where the ENUM client presents a list of possible URLs to the end
user for his or her choice, it MUST do so in the sequence defined by
the ORDER and PREFERENCE/PRIORITY values specified by the Registrant.

However, a Registrant SHOULD place into his or her zone only contacts
that he or she is willing to support; even those with the worst ORDER
and PREFERENCE/PRIORITY values MAY be selected by an end user.

4.3. Use of Order and Preference Fields

NAPTRs in ENUM zones that hold incorrect ORDER values can cause major
problems. [RFC3403] highlights that having both ORDER and
PREFERENCE/PRIORITY fields is a historical artifact of the NAPTR
resource record type. It is reasonable to have a common default
value for the ORDER field, relying on the PREFERENCE/PRIORITY field
to indicate the preferred sort.

Conroy & Fujiwara Informational [Page 13]

RFC 5483 ENUM Experiences March 2009

We have noticed a number of ENUM domains with NAPTRs that have
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, or
simply a difference of interpretation of the standards.

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; not the
other way around.

To avoid these common interoperability issues, it is recommended that
ENUM NAPTRs SHOULD hold a default value in their ORDER field.

4.4. NAPTRs with Identical ORDER/PRIORITY Values

From experience, it has been learned that 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 normally occur.

Such a condition indicates that these NAPTRs are truly identical in
priority and that there is no preference between the services these
NAPTRs offer. Implementers SHOULD NOT assume that the DNS will
deliver NAPTRs within an RRSet in a particular sequence.

Multiple NAPTRs with identical ORDER and identical PREFERENCE/
PRIORITY field values SHOULD NOT be provisioned into an RRSet unless
the intent is that these NAPTRs are truly identical in priority and
there is no preference between them.

Some ENUM client implementations have considered this case to be an
error and have rejected such duplicates entirely. Others have
attempted to further randomise the order in which such duplicates are
processed. Thus, use of such duplicate NAPTRs is unwise, as client
implementations exist that will behave in different ways.

4.4.1. Compound NAPTRs and Implicit ORDER/REFERENCE Values

With [RFC3761], it is possible to have more than one Enumservice
associated with a single NAPTR. These Enumservices share the same
Regexp field and so generate the same URI. Such a "compound" NAPTR
could well be used to indicate a mobile phone that supports both
"voice:tel" and "sms:tel" Enumservices. The Services field in that
case would be "E2U+voice:tel+sms:tel".

Conroy & Fujiwara Informational [Page 14]

RFC 5483 ENUM Experiences March 2009

A compound NAPTR can be treated as a set of NAPTRs that each hold a
single Enumservice. These reconstructed NAPTRs share the same ORDER
and PREFERENCE/PRIORITY field values but should be treated as if each
had a logically different priority. In this case, the reconstructed
NAPTR holding the leftmost Enumservice within the compound NAPTR has
the best priority, and the reconstructed NAPTR holding the rightmost
Enumservice has the worst priority in this set.

To avoid indeterminate behaviour, it is recommended that ENUM clients
SHOULD process the Enumservices within a compound NAPTR in a left-to-
right sequence. ENUM provisioning systems SHOULD assume that such a
processing order will be used and provision the Enumservices within a
compound NAPTR accordingly.

4.5. Processing Order Value across Domains

Using a different ORDER field value in different domains is
unimportant for most queries. However, DDDS includes a mechanism for
continuing a search for NAPTRs in another domain by including a
reference to that other domain in a "non-terminal" NAPTR. The
treatment of non-terminal NAPTRs is covered in the next section. If
they are supported, then the way that ORDER and PREFERENCE/PRIORITY
field values are processed is affected.

Two main questions remain from the specifications of DDDS and
[RFC3761]:

o If there is a different (lower) ORDER field value in a domain
referred to by a non-terminal NAPTR, then does this mean that the
ENUM client discards any remaining NAPTRs in the referring RRSet?

o Conversely, if the domain referred to by a non-terminal NAPTR
contains entries that only have a higher ORDER field value, then
does the ENUM client ignore those NAPTRs in the referenced domain?

Whilst one interpretation of [RFC3761] 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.

In keeping with the interpretation made so far, ENUM implementations
MUST consider the ORDER and PREFERENCE/PRIORITY values only within
the context of the domain currently being processed in an ENUM query.
These values MUST be discarded when processing other RRSets in the
query.

Conroy & Fujiwara Informational [Page 15]

RFC 5483 ENUM Experiences March 2009

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 holds, as its target, 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
referenced 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 form 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" Private
Branch Exchange (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, some existing ENUM clients do not support non-terminal
NAPTRs and ignore them if received.

To avoid interoperability problems, some kind of acceptable advice is
needed on non-terminal NAPTRs. As current support is limited, non-
terminal NAPTRs SHOULD NOT be used in ENUM unless it is clear that
all of the ENUM clients this environment supports can process these.

Conroy & Fujiwara Informational [Page 16]

RFC 5483 ENUM Experiences March 2009

5.2. Non-Terminal NAPTRs - Considerations

The following specific issues need to be considered if non-terminal
NAPTRs are to be supported in a particular environment. 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 RRSet refers to the
NAPTRs contained in another domain. The NAPTRs in the domain
referred to by the non-terminal NAPTR may have a different ORDER
value from that in the referring non-terminal NAPTR. See Section 4.5
for details.

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, a "non-terminal" or referential loop
is implied. An implementation MAY treat a chain of more than 5
domains traversed during a single ENUM query as an indication that a
self-referential loop has been entered.

There are many techniques that can be used to detect such a loop, but
the simple approach of counting the number of domains queried in the
current ENUM query suffices.

Where a loop has been detected, processing SHOULD continue at the
next NAPTR in the referring domain (i.e., after the non-terminal
NAPTR that included the reference that triggered the loop detection).

5.2.3. Field Content in Non-Terminal NAPTRs

The set of specifications defining DDDS and its applications are
complex and multi-layered. This reflects the flexibility that the
system provides but does mean that some of the specifications need
clarification as to their interpretation, particularly where non-
terminal rules are concerned.

Conroy & Fujiwara Informational [Page 17]

RFC 5483 ENUM Experiences March 2009

5.2.3.1. Flags Field Content with Non-Terminal NAPTRs

Section 2.4.1 of [RFC3761] states that the only flag character valid
for use with the "E2U" DDDS Application is 'u'. The flag 'u' is
defined (in Section 4.3 of [RFC3404]) thus: 'The "u" flag means that
the output of the Rule is a URI'.

Section 2.4.1 of [RFC3761] 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 [RFC3404]. One
could well argue that this is a feature potentially common to all
DDDS Applications, and so might have been specified in [RFC3402] or
[RFC3403].

The Flags field will be empty in non-terminal NAPTRs encountered in
ENUM processing. ENUM does not have any other way to indicate a non-
terminal NAPTR.

5.2.3.2. Services Field Content with Non-Terminal NAPTRs

Furthermore, [RFC3761] 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 a non-terminal NAPTR cannot hold a valid
(non-empty) Services field when used in ENUM. Section 2.4.2 of
[RFC3761] specifies the syntax for this field content and requires at
least one element of type <servicespec> (i.e., at least one
Enumservice identifier). Given that there cannot be a non-terminal
Enumservice (and so no such Registered Enumservice identifier), this
syntax cannot be met with a non-terminal NAPTR; there are no non-
terminal Enumservices to put into this field.

A reasonable interpretation of the specifications is that for a non-
terminal NAPTR, 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.

It is expected that future revisions of the ENUM standard will
clarify this text, making this interpretation plain. This was the
intent of the current standard, and the intent will be made explicit
in its revision.

Conroy & Fujiwara Informational [Page 18]

RFC 5483 ENUM Experiences March 2009

In keeping with existing implementations, in a non-terminal NAPTR
encountered in an ENUM query, the Services field SHOULD be empty, and
clients SHOULD ignore any content it contains.

Of course, such non-terminal NAPTRs with an empty Services field are
not specific to any DDDS Application. Thus, other means must be used
to ensure a non-terminal NAPTR that is intended only for a particular
DDDS Application cannot be encountered during a lookup for another
DDDS Application (for example, by ensuring that the same domain is
not used to host NAPTRs for more than one such DDDS Application).

5.2.3.3. Regular Expression and Replacement Field Content with Non-
Terminal NAPTRs

The descriptive text in Section 4.1 of [RFC3403] 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.

[RFC3403] 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,
[RFC3403] does not specify which is used with terminal and non-
terminal rules.

The descriptive text of Section 4.1 of [RFC3403] 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
should 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.

Conroy & Fujiwara Informational [Page 19]

RFC 5483 ENUM Experiences March 2009

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 includes specifying 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 what is the expected
output). ENUM (as specified in Section 2.4.1 of [RFC3761]) uses 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 implementations. 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. Generating an ENUM domain name from the Regexp field is
difficult at best and impossible for the general case of a variable-
length telephone number, or one that has more than 9 digits. Thus,
it is proposed that when the ENUM specification is updated, this
option is deprecated, and using the Regexp field for non-terminal
ENUM NAPTRs is prohibited.

In keeping with current implementations, the target domain of a non-
terminal ENUM NAPTR MUST be placed in the (non-empty) Replacement
field. This field MUST be interpreted as holding the domain name
that forms the next key output from this non-terminal rule.
Conversely, the Regexp field MUST be empty in a non-terminal NAPTR
encountered in ENUM processing, and ENUM clients MUST ignore its
content.

6. Backwards Compatibility

6.1. Services Field Syntax

[RFC3761] is the current standard for the syntax for NAPTRs
supporting the ENUM DDDS Application. This obsoletes the original
specification that was given in [RFC2916]. RFC 3761 made a change to
the syntax of the Services field of the NAPTR that reflects a
refinement of the concept of ENUM processing.

Conroy & Fujiwara Informational [Page 20]

RFC 5483 ENUM Experiences March 2009

As defined in [RFC3403], there is now a single identifier that
indicates the DDDS Application. In the obsolete specification
[RFC2915], there were zero or more "Resolution Service" identifiers
(the equivalent of the DDDS Application). The same identifier string
for the DDDS identifier or the Resolution Service is defined in both
the [RFC3761] and [RFC2916] specifications: "E2U".

Also, [RFC3761] 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. [RFC3761] 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. [RFC2916] specified that the protocol sub-
field was the leftmost, followed by the '+' delimiter, in turn
followed by the "E2U" resolution service tag.

[RFC2915] and [RFC2916] have been obsoleted by [RFC3401] - [RFC3404]
and by [RFC3761]. However, [RFC3824] 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. 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.

To facilitate this method, IANA should reject any request to register
an Enumservice with the label "E2U".

To summarise, ENUM clients MUST support ENUM NAPTRs according to
[RFC3761] syntax. ENUM clients SHOULD also support ENUM NAPTRs
according to the obsolete syntax of [RFC2916]; there are still zones
that hold "old" syntax NAPTRs. ENUM zones MUST NOT be provisioned
with NAPTRs according to the obsolete form, and MUST be provisioned
with NAPTRs in which the Services field is according to [RFC3761].

7. Collected Implications for ENUM Provisioning

ENUM NAPTRs SHOULD NOT include characters outside the printable US-
ASCII equivalent range (U+0020 to U+007E) unless it is clear that all
ENUM clients they are designed to support will be able to process
such characters correctly. If ENUM zone provisioning systems require
non-ASCII characters, these systems SHOULD encode the non-ASCII data
to emit only US-ASCII characters by applying the appropriate

Conroy & Fujiwara Informational [Page 21]

RFC 5483 ENUM Experiences March 2009

mechanism ([RFC3492], [RFC3987]). Non-printable characters SHOULD
NOT be used, as ENUM clients may need to present NAPTR content in a
human-readable form.

The case-sensitivity flag ('i') is inappropriate for ENUM, and SHOULD
NOT be provisioned into the Regexp field of E2U NAPTRs.

ENUM zone provisioning systems SHOULD use '!' (U+0021) as their
Regexp delimiter character.

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 [RFC3402]
(i.e., "\!", U+005C U+0021). Note that when a NAPTR resource record
is entered in DNS master file syntax, the backslash itself must be
escaped using a second backslash.

If present in the ERE sub-field of an ENUM NAPTR, the literal
character '+' MUST be escaped as "\+" (i.e. U+005C U+002B). Note
that, as always, when a NAPTR resource record is entered in DNS
master file syntax, the backslash itself must be escaped using a
second backslash.

The Registrant and the ENUM zone provisioning system he or she uses
SHOULD NOT rely on ENUM clients solely taking account of the value of
the ORDER and the PREFERENCE/PRIORITY fields in ENUM NAPTRs. Thus, a
Registrant SHOULD place into his or her zone only contacts that he or
she is willing to support; even those with the worst ORDER and
PREFERENCE/PRIORITY values MAY be selected by an end user.

Many apparent mistakes in ORDER and PREFERENCE/PRIORITY values have
been detected in provisioned ENUM zones. To avoid these common
interoperability issues, provisioning systems SHOULD NOT use
different ORDER field values for NAPTRs in a Resource Record Set
(RRSet). To generalise, all ENUM NAPTRs SHOULD hold a default value
in their ORDER field. A value of "100" is recommended, as it seems
to be used in most provisioned domains.

Multiple NAPTRs with identical ORDER and identical PREFERENCE/
PRIORITY field values SHOULD NOT be provisioned into an RRSet unless
the intent is that these NAPTRs are truly identical and there is no
preference between them. Implementers SHOULD NOT assume that the DNS
will deliver NAPTRs within an RRSet in a particular sequence.

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.

Conroy & Fujiwara Informational [Page 22]

RFC 5483 ENUM Experiences March 2009

ENUM zone provisioning systems SHOULD assume that, once a non-
terminal NAPTR has been selected for processing, the ORDER field
value in a domain referred to by that non-terminal NAPTR will be
considered only within the context of that referenced domain (i.e.,
the ORDER value will be used only to sort within the current RRSet
and will not be used in the processing of NAPTRs in any other RRSet).

Whilst this client behaviour is non-compliant, ENUM provisioning
systems and their users should be aware that some ENUM clients have
been detected with poor (or no) support for non-trivial ERE sub-field
expressions.

ENUM provisioning systems SHOULD be cautious in the use of multiple
back-reference patterns in the Repl sub-field of NAPTRs they
provision. Some clients have limited buffer space for character
expansion when generating URIs (see also Section 3). These
provisioning systems SHOULD check the back-reference replacement
patterns they use, ensuring that regular expression processing will
not produce excessive-length URIs.

As current support is limited, non-terminal NAPTRs SHOULD NOT be
provisioned in ENUM zones unless it is clear that all ENUM clients
that this environment supports can process these.

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.

In a non-terminal NAPTR encountered in an ENUM query (i.e., one with
an empty Flags field), the Services field SHOULD be empty.

A non-terminal NAPTR MUST include its target domain in the (non-
empty) Replacement field. This field MUST be interpreted as holding
the domain name that forms the next key output from this non-terminal
rule. The Regexp field MUST be empty in a non-terminal NAPTR
intended to be encountered during an ENUM query.

ENUM zones MUST NOT be provisioned with NAPTRs according to the
obsolete form, and MUST be provisioned with NAPTRs in which the
Services field is according to [RFC3761].

8. Collected Implications for ENUM Clients

ENUM clients SHOULD NOT discard NAPTRs in which they detect
characters outside the US-ASCII printable range (0x20 to 0x7E
hexadecimal).

Conroy & Fujiwara Informational [Page 23]

RFC 5483 ENUM Experiences March 2009

ENUM clients MAY discard NAPTRs that have octets in the Flags,
Services, or Regexp fields that have byte values outside the US-ASCII
equivalent range (i.e., byte values above 0x7F). Clients MUST be
ready to encounter NAPTRs with such values without failure.

ENUM clients SHOULD NOT assume that the delimiter is the last
character of the Regexp field.

Unless they are sure that in their environment this is the case,
in general an ENUM client may still encounter NAPTRs that have
been provisioned with a following 'i' (case-insensitive) flag,
even though that flag has no effect at all in an ENUM scenario.

ENUM clients SHOULD discard NAPTRs that have more or less than 3
unescaped instances of the delimiter character within the Regexp
field.

In the spirit of being liberal with what it will accept, if the
ENUM client is sure how the Regexp field should be interpreted,
then it may choose to process the NAPTR even in the face of an
incorrect number of unescaped delimiter characters. If it is not
clear how the Regexp field should be interpreted, then the client
must discard the NAPTR.

Where the ENUM client presents a list of possible URLs to the end
user for his or her choice, it MAY present all NAPTRs -- not just the
ones with the highest currently unprocessed ORDER field value. The
client SHOULD keep to the ORDER and PREFERENCE/PRIORITY values
specified by the Registrant.

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 might have done this already).

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.

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.

ENUM clients MUST be ready to process NAPTRs that use a different
character from '!' as their Regexp Delimiter without failure.

Conroy & Fujiwara Informational [Page 24]

RFC 5483 ENUM Experiences March 2009

ENUM clients MUST be ready to process NAPTRs that have non-trivial
patterns in their ERE sub-field values without failure.

ENUM clients MUST be ready to process NAPTRs with a DDDS Application
identifier other than 'E2U' without failure.

ENUM clients MUST be ready to process NAPTRs with many copies of
back-reference patterns within the Repl sub-field without failure
(see also Section 3).

If a NAPTR is discarded, this SHOULD NOT cause the whole ENUM query
to terminate and processing SHOULD continue with the next NAPTR in
the returned Resource Record Set (RRSet).

When an ENUM client encounters a compound NAPTR (i.e., one containing
more than one Enumservice) and cannot process or cannot recognise one
of the Enumservices within it, that ENUM client SHOULD ignore this
Enumservice 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. These conditions SHOULD NOT be
considered errors.

ENUM clients MUST support ENUM NAPTRs according to [RFC3761] syntax.
ENUM clients SHOULD also support ENUM NAPTRs according to the
obsolete syntax of [RFC2916]; there are still zones that hold "old"
syntax NAPTRs.

8.1. Non-Terminal NAPTR Processing

ENUM clients MUST be ready to process NAPTRs with an empty Flags
field ("non-terminal" NAPTRs) without failure. More generally, non-
terminal NAPTR processing SHOULD be implemented, but ENUM clients MAY
discard non-terminal NAPTRs they encounter.

ENUM clients SHOULD ignore any content of the Services field when
encountering a non-terminal NAPTR with an empty Flags field.

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. If
present in a non-terminal NAPTR, a non-empty Regexp field MUST be
ignored by ENUM clients.

Conroy & Fujiwara Informational [Page 25]

RFC 5483 ENUM Experiences March 2009

If a problem is detected when processing an ENUM query across
multiple domains (by following non-terminal NAPTR references), then
the ENUM query SHOULD NOT be abandoned, but instead processing SHOULD
continue at the next NAPTR after the non-terminal NAPTR that referred
to the domain in which the problem would have occurred.

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).

ENUM clients MAY consider a chain of more than 5 "non-terminal"
NAPTRs traversed in a single ENUM query as an indication that a
referential loop has been entered.

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
referential 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.

9. Security Considerations

In addition to the security implications of recommendations in this
document, 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 2, Section 4, and Section 6 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 Informational [Page 26]

RFC 5483 ENUM Experiences March 2009

As mentioned in Section 3, ENUM uses regular expressions to generate
URIs. Though it is a standard feature of DDDS, use of "non-greedy"
regular expressions with multiple back-reference patterns in the Repl
sub-field does create the potential for buffer-overrun attacks.
Provisioning system designers SHOULD be aware of this and SHOULD
limit the repeated use of back-reference replacement patterns.
Conversely, ENUM client implementers SHOULD avoid using fixed
character buffers when generating URIs from Repl sub-fields that
include Back-reference patterns, and MUST avoid failure in the case
of buffer exhaustion.

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 [ETSI-TS102172] 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 rigorous framework on which to build.
This document reflects how complex a system it is: without the
intricacy of [RFC3401] - [RFC3404] and the work that went into them,
it could have been very difficult to ensure interoperability.

11. References

11.1. Normative References

[E.164] ITU-T, "The International Public Telecommunication Number
Plan", Recommendation E.164, February 2005.

[IEEE.1003-2.1992]
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.

[RFC1034] Mockapetris, P., "Domain names - concepts and facilities",
STD 13, RFC 1034, November 1987.

Conroy & Fujiwara Informational [Page 27]

RFC 5483 ENUM Experiences March 2009

[RFC1035] Mockapetris, P., "Domain names - implementation and
specification", STD 13, RFC 1035, November 1987.

[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, March 1997.

[RFC3402] Mealling, M., "Dynamic Delegation Discovery System (DDDS)
Part Two: The Algorithm", RFC 3402, October 2002.

[RFC3403] Mealling, M., "Dynamic Delegation Discovery System (DDDS)
Part Three: The Domain Name System (DNS) Database",
RFC 3403, October 2002.

[RFC3404] Mealling, M., "Dynamic Delegation Discovery System (DDDS)
Part Four: The Uniform Resource Identifiers (URI)",
RFC 3404, October 2002.

[RFC3405] Mealling, M., "Dynamic Delegation Discovery System (DDDS)
Part Five: URI.ARPA Assignment Procedures", BCP 65,
RFC 3405, October 2002.

[RFC3490] Faltstrom, P., Hoffman, P., and A. Costello,
"Internationalizing Domain Names in Applications (IDNA)",
RFC 3490, March 2003.

[RFC3491] Hoffman, P. and M. Blanchet, "Nameprep: A Stringprep
Profile for Internationalized Domain Names (IDN)",
RFC 3491, March 2003.

[RFC3492] Costello, A., "Punycode: A Bootstring encoding of Unicode
for Internationalized Domain Names in Applications
(IDNA)", RFC 3492, March 2003.

[RFC3629] Yergeau, F., "UTF-8, a transformation format of ISO
10646", STD 63, RFC 3629, November 2003.

[RFC3761] Faltstrom, P. and M. Mealling, "The E.164 to Uniform
Resource Identifiers (URI) Dynamic Delegation Discovery
System (DDDS) Application (ENUM)", RFC 3761, April 2004.

[RFC3966] Schulzrinne, H., "The tel URI for Telephone Numbers",
RFC 3966, December 2004.

[RFC3986] Berners-Lee, T., Fielding, R., and L. Masinter, "Uniform
Resource Identifier (URI): Generic Syntax", STD 66,
RFC 3986, January 2005.

Conroy & Fujiwara Informational [Page 28]

RFC 5483 ENUM Experiences March 2009

[RFC3987] Duerst, M. and M. Suignard, "Internationalized Resource
Identifiers (IRIs)", RFC 3987, January 2005.

11.2. Informative References

[ASCII] American National Standards Institute, "Coded Character
Set - 7-bit American Standard Code for Information
Interchange", ANSI X3.4, 1986.

[ETSI-TS102172]
ETSI, "Minimum Requirements for Interoperability of
European ENUM Implementations", ETSI TS 102 172,
October 2004.

[RFC2915] Mealling, M. and R. Daniel, "The Naming Authority Pointer
(NAPTR) DNS Resource Record", RFC 2915, September 2000.

[RFC2916] Faltstrom, P., "E.164 number and DNS", RFC 2916,
September 2000.

[RFC3401] Mealling, M., "Dynamic Delegation Discovery System (DDDS)
Part One: The Comprehensive DDDS", RFC 3401, October 2002.

[RFC3824] Peterson, J., Liu, H., Yu, J., and B. Campbell, "Using
E.164 numbers with the Session Initiation Protocol (SIP)",
RFC 3824, June 2004.

Conroy & Fujiwara Informational [Page 29]

RFC 5483 ENUM Experiences March 2009

Authors' Addresses

Lawrence Conroy
Roke Manor Research
Roke Manor
Old Salisbury Lane
Romsey
United Kingdom

Phone: +44-1794-833666
EMail: [email protected]
URI: http://www.sienum.co.uk

Kazunori Fujiwara
Japan Registry Services Co., Ltd.
Chiyoda First Bldg. East 13F
3-8-1 Nishi-Kanda Chiyoda-ku
Tokyo 101-0165
JAPAN

EMail: [email protected]
URI: http://jprs.co.jp/en/

Conroy & Fujiwara Informational [Page 30]