self delimiting, or be delimited by self delimiting tokens.
Comments and LWSP are not used as delimiters, except for the
following cases, where LWSP may be inserted according to RFC
822 rules.
- Around the ":" in all headers
- EBNF.labelled-integer
- EBNF.object-identifier
- EBNF.encoded-info
RFC822 folding rules are applied to all headers.
3.1.2. ASN.1
An element is referred to with the following syntax, defined in EBNF:
element = service "." definition *( "." definition )
service = "IPMS" / "MTS" / "MTA"
definition = identifier / context
identifier = ALPHA *< ALPHA or DIGIT or "-" >
context = "[" 1*DIGIT "]"
The EBNF.service keys are shorthand for the following service
specifications:
IPMS IPMSInformationObjects defined in Annex E of X.420 / ISO
10021-7.
MTS MTSAbstractService defined in Section 9 of X.411 / ISO
10021-4.
MTA MTAAbstractService defined in Section 13 of X.411 / ISO
10021-4.
The first EBNF.identifier identifies a type or value key in the
context of the defined service specification. Subsequent
EBNF.identifiers identify a value label or type in the context of the
first identifier (SET or SEQUENCE). EBNF.context indicates a context
tag, and is used where there is no label or type to uniquely identify
a component. The special EBNF.identifier keyword "value" is used to
denote an element of a sequence.
For example, IPMS.Heading.subject defines the subject element of the
IPMS heading. The same syntax is also used to refer to element
values. For example,
MTS.EncodedInformationTypes.[0].g3Fax refers to a value of
MTS.EncodedInformationTypes.[0] .
3.2. ASCII and IA5
A gateway will interpret all IA5 as ASCII. Thus, mapping between
these forms is conceptual.
3.3. Standard Types
There is a need to convert between ASCII text, and some of the types
defined in ASN.1 [CCITT/ISO88d]. For each case, an EBNF syntax
definition is given, for use in all of this specification, which
leads to a mapping between ASN.1, and an EBNF construct. All EBNF
syntax definitions of ASN.1 types are in lower case, whereas ASN.1
types are referred to with the first letter in upper case. Except as
noted, all mappings are symmetrical.
3.3.1. Boolean
Boolean is encoded as:
boolean = "TRUE" / "FALSE"
3.3.2. NumericString
NumericString is encoded as:
numericstring = *DIGIT
3.3.3. PrintableString
PrintableString is a restricted IA5String defined as:
printablestring = *( ps-char )
ps-restricted-char = 1DIGIT / 1ALPHA / " " / "'" / "+"
/ "," / "-" / "." / "/" / ":" / "=" / "?"
ps-delim = "(" / ")"
ps-char = ps-delim / ps-restricted-char
This can be used to represent real printable strings in EBNF.
3.3.4. T.61String
In cases where T.61 strings are only used for conveying human
interpreted information, the aim of a mapping is to render the
characters appropriately in the remote character set, rather than to
maximise reversibility. For these cases, the mappings to IA5 defined
in CCITT Recommendation X.408 (1988) shall be used [CCITT/ISO88a].
These will then be encoded in ASCII.
There is also a need to represent Teletex Strings in ASCII, for some
aspects of O/R Address. For these, the following encoding is used:
teletex-string = *( ps-char / t61-encoded )
t61-encoded = "{" 1* t61-encoded-char "}"
t61-encoded-char = 3DIGIT
Common characters are mapped simply. Other octets are mapped using a
quoting mechanism similar to the printable string mechanism. Each
octet is represented as 3 decimal digits.
There are a number of places where a string may have a Teletex and/or
Printable String representation. The following BNF is used to
represent this.
teletex-and-or-ps = [ printablestring ] [ "*" teletex-string ]
The natural mapping is restricted to EBNF.ps-char, in order to make
the full BNF easier to parse.
3.3.5. UTCTime
Both UTCTime and the RFC822 822.date-time syntax contain: Year
(lowest two digits), Month, Day of Month, hour, minute, second
(optional), and Timezone. 822.date-time also contains an optional
day of the week, but this is redundant. Therefore a symmetrical
mapping can be made between these constructs.
Note:
In practice, a gateway will need to parse various illegal
variants on 822.date-time. In cases where 822.date-time
cannot be parsed, it is recommended that the derived UTCTime
is set to the value at the time of translation.
When mapping to X.400, the UTCTime format which specifies the
timezone offset shall be used.
When mapping to RFC822, the 822.date-time format shall include a
numeric timezone offset (e.g., +0000).
When mapping time values, the timezone shall be preserved as
specified. The date shall not be normalised to any other timezone.
3.3.6. Integer
A basic ASN.1 Integer will be mapped onto EBNF.numericstring. In
many cases ASN.1 will enumerate Integer values or use ENUMERATED. An
EBNF encoding labelled-integer is provided. When mapping from EBNF to
ASN.1, only the integer value is mapped, and the associated text is
discarded. When mapping from ASN.1 to EBNF, addition of an
appropriate text label is strongly encouraged.
labelled-integer ::= [ key-string ] "(" numericstring ")"
key-string = *key-char
key-char = <a-z, A-Z, 0-9, and "-">
3.3.7. Object Identifier
Object identifiers are represented in a form similar to that given in
ASN.1. The order is the same as for ASN.1 (big-endian). The numbers
are mandatory, and used when mapping from the ASCII to ASN.1. The
key-strings are optional. It is recommended that as many strings as
possible are generated when mapping from ASN.1 to ASCII, to
facilitate user recognition.
object-identifier ::= oid-comp object-identifier
| oid-comp
oid-comp ::= [ key-string ] "(" numericstring ")"
An example representation of an object identifier is:
joint-iso-ccitt(2) mhs (6) ipms (1) ep (11) ia5-text (0)
or
(2) (6) (1)(11)(0)
3.4. Encoding ASCII in Printable String
Some information in RFC822 is represented in ASCII, and needs to be
mapped into X.400 elements encoded as printable string. For this
reason, a mechanism to represent ASCII encoded as PrintableString is
needed.
A structured subset of EBNF.printablestring is now defined. This
shall be used to encode ASCII in the PrintableString character set.
ps-encoded = *( ps-restricted-char / ps-encoded-char )
ps-encoded-char = "(a)" ; (@)
/ "(p)" ; (%)
/ "(b)" ; (!)
/ "(q)" ; (")
/ "(u)" ; (_)
/ "(l)" ; "("
/ "(r)" ; ")"
/ "(" 3DIGIT ")"
The 822.3DIGIT in EBNF.ps-encoded-char must have range 0-127, and is
interpreted in decimal as the corresponding ASCII character. Special
encodings are given for: at sign (@), percent (%), exclamation
mark/bang (!), double quote ("), underscore (_), left bracket ((),
and right bracket ()). These characters, with the exception of round
brackets, are not included in PrintableString, but are common in RFC
822 addresses. The abbreviations will ease specification of RFC822
addresses from an X.400 system. These special encodings shall be
interpreted in a case insensitive manner, but always generated in
lower case.
A reversible mapping between PrintableString and ASCII can now be
defined. The reversibility means that some values of printable
string (containing round braces) cannot be generated from ASCII.
Therefore, this mapping must only be used in cases where the
printable strings may only be derived from ASCII (and will therefore
have a restricted domain). For example, in this specification, it is
only applied to a Domain Defined Attribute which will have been
generated by use of this specification and a value such as "(" would
not be possible.
To encode ASCII as PrintableString, the EBNF.ps-encoded syntax is
used, with all EBNF.ps-restricted-char mapped directly. All other
822.CHAR are encoded as EBNF.ps-encoded-char.
To encode PrintableString as ASCII, parse PrintableString as
EBNF.ps-encoded, and then reverse the previous mapping. If the
PrintableString cannot be parsed, then the mapping is being applied
in to an inappropriate value, and an error shall be given to the
procedure doing the mapping. In some cases, it may be preferable to
pass the printable string through unaltered.
Some examples are now given. Note the arrows which indicate
asymmetrical mappings:
PrintableString ASCII
'a demo.' <-> 'a demo.'
foo(a)bar <-> foo@bar
(q)(u)(p)(q) <-> "_%"
(a) <-> @
(A) -> @
(l)a(r) <-> (a)
(126) <-> ~
( -> (
(l) <-> (
Chapter 4 - Addressing
Addressing is probably the trickiest problem of an X.400 <-> RFC822
gateway. Therefore it is given a separate chapter. This chapter, as
a side effect, also defines a textual representation of an X.400 O/R
Address.
Initially we consider an address in the (human) mail user sense of
"what is typed at the mailsystem to reference a mail user". A basic
RFC822 address is defined by the EBNF EBNF.822-address:
822-address = [ route ] addr-spec
In an 822-MTS protocol, the originator and each recipient are
considered to be defined by such a construct. In an RFC822 header,
the EBNF.822-address is encapsulated in the 822.address syntax rule,
and there may also be associated comments. None of this extra
information has any semantics, other than to the end user.
The basic X.400 O/R Address, used by the MTS for routing, is defined
by MTS.ORAddress. In IPMS, the MTS.ORAddress is encapsulated within
IPMS.ORDescriptor.
It can be seen that RFC822 822.address must be mapped with
IPMS.ORDescriptor, and that RFC822 EBNF.822-address must be mapped
with MTS.ORAddress.
4.1. A textual representation of MTS.ORAddress
MTS.ORAddress is structured as a set of attribute value pairs. It is
clearly necessary to be able to encode this in ASCII for gatewaying
purposes. All components shall be encoded, in order to guarantee
return of error messages, and to optimise third party replies.
4.2. Basic Representation
An O/R Address has a number of structured and unstructured
attributes. For each unstructured attribute, a key and an encoding
is specified. For structured attributes, the X.400 attribute is
mapped onto one or more attribute value pairs. For domain defined
attributes, each element of the sequence will be mapped onto a triple
(key and two values), with each value having the same encoding. The
attributes are as follows, with 1984 attributes given in the first
part of the table. For each attribute, a reference is given,
consisting of the relevant sections in X.402 / ISO 10021-2, and the
extension identifier for 88 only attributes:
Attribute (Component) Key Enc Ref Id
84/88 Attributes
MTS.CountryName C P 18.3.3
MTS.AdministrationDomainName ADMD P 18.3.1
MTS.PrivateDomainName PRMD P 18.3.21
MTS.NetworkAddress X121 N 18.3.7
MTS.TerminalIdentifier T-ID P 18.3.23
MTS.OrganizationName O P/T 18.3.9
MTS.OrganizationalUnitNames.value OU P/T 18.3.10
MTS.NumericUserIdentifier UA-ID N 18.3.8
MTS.PersonalName PN P/T 18.3.12
MTS.PersonalName.surname S P/T 18.3.12
MTS.PersonalName.given-name G P/T 18.3.12
MTS.PersonalName.initials I P/T 18.3.12
MTS.PersonalName
.generation-qualifier GQ P/T 18.3.12
MTS.DomainDefinedAttribute.value DD P/T 18.1
88 Attributes
MTS.CommonName CN P/T 18.3.2 1
MTS.TeletexCommonName CN P/T 18.3.2 2
MTS.TeletexOrganizationName O P/T 18.3.9 3
MTS.TeletexPersonalName PN P/T 18.3.12 4
MTS.TeletexPersonalName.surname S P/T 18.3.12 4
MTS.TeletexPersonalName.given-name G P/T 18.3.12 4
MTS.TeletexPersonalName.initials I P/T 18.3.12 4
MTS.TeletexPersonalName
.generation-qualifier GQ P/T 18.3.12 4
MTS.TeletexOrganizationalUnitNames
.value OU P/T 18.3.10 5
MTS.TeletexDomainDefinedAttribute
.value DD P/T 18.1 6
MTS.PDSName PD-SERVICE P 18.3.11 7
MTS.PhysicalDeliveryCountryName PD-C P 18.3.13 8
MTS.PostalCode PD-CODE P 18.3.19 9
MTS.PhysicalDeliveryOfficeName PD-OFFICE P/T 18.3.14 10
MTS.PhysicalDeliveryOfficeNumber PD-OFFICE-NUM P/T 18.3.15 11
MTS.ExtensionORAddressComponents PD-EXT-ADDRESS P/T 18.3.4 12
MTS.PhysicalDeliveryPersonName PD-PN P/T 18.3.17 13
MTS.PhysicalDeliveryOrganizationName PD-O P/T 18.3.16 14
MTS.ExtensionPhysicalDelivery
AddressComponents PD-EXT-DELIVERY P/T 18.3.5 15
MTS.UnformattedPostalAddress PD-ADDRESS P/T 18.3.25 16
MTS.StreetAddress PD-STREET P/T 18.3.22 17
MTS.PostOfficeBoxAddress PD-BOX P/T 18.3.18 18
MTS.PosteRestanteAddress PD-RESTANTE P/T 18.3.20 19
MTS.UniquePostalName PD-UNIQUE P/T 18.3.26 20
MTS.LocalPostalAttributes PD-LOCAL P/T 18.3.6 21
MTS.ExtendedNetworkAddress
.e163-4-address.number NET-NUM N 18.3.7 22
MTS.ExtendedNetworkAddress
.e163-4-address.sub-address NET-SUB N 18.3.7 22
MTS.ExtendedNetworkAddress
.psap-address NET-PSAP X 18.3.7 22
MTS.TerminalType T-TY I 18.3.24 23
The following keys identify different EBNF encodings, which are
associated with the ASCII representation of MTS.ORAddress.
Key Encoding
P printablestring
N numericstring
T teletex-string
P/T teletex-and-or-ps
I labelled-integer
X presentation-address
The BNF for presentation-address is taken from the specification "A
String Encoding of Presentation Address" [Kille89a].
In most cases, the EBNF encoding maps directly to the ASN.1 encoding
of the attribute. There are a few exceptions. In cases where an
attribute can be encoded as either a PrintableString or NumericString
(Country, ADMD, PRMD), either form is mapped into the BNF. When
generating ASN.1, the NumericString encoding shall be used if the
string contains only digits.
There are a number of cases where the P/T (teletex-and-or-ps)
representation is used. Where the key maps to a single attribute,
this choice is reflected in the encoding of the attribute (attributes
10-21). For most of the 1984 attributes and common name, there is a
printablestring and a teletex variant. This pair of attributes is
mapped onto the single component here. This will give a clean
mapping for the common cases where only one form of the name is used.
Recently, ISO has undertaken work to specify a string form of O/R
Address [CCITT/ISO91a]. This has specified a number of string
keywords for attributes. As RFC1148 was an input to this work, many
of the keywords are the same. To increase compatability, the
following alternative values shall be recognised when mapping from
RFC822 to X.400. These shall not be generated when mapping from
X.400 to RFC822.
Keyword Alternative
ADMD A
PRMD P
GQ Q
X121 X.121
UA-ID N-ID
PD-OFFICE-NUMBER PD-OFFICE NUMBER
When mapping from RFC822 to X.400, the keywords: OU1, OU2, OU3, and
OU4, shall be recognised. If these are present, no keyword OU
shall be present. These will be treated as ordered values of OU.
4.2.1. Encoding of Personal Name
Handling of Personal Name and Teletex Personal Name based purely on
the EBNF.standard-type syntax defined above is likely to be clumsy.
It seems desirable to utilise the "human" conventions for encoding
these components. A syntax is defined, which is designed to provide
a clean encoding for the common cases of O/R Address specification
where:
1. There is no generational qualifier
2. Initials contain only letters
3. Given Name does not contain full stop ("."), and is at least
two characters long.
4. Surname does not contain full stop in the first two
characters.
5 If Surname is the only component, it does not contain full
stop.
The following EBNF is defined:
encoded-pn = [ given "." ] *( initial "." ) surname
given = 2*<ps-char not including ".">
initial = ALPHA
surname = printablestring
This is used to map from any string containing only printable string
characters to an O/R address personal name. To map from a string to
O/R Address components, parse the string according to the EBNF. The
given name and surname are assigned directly. All EBNF.initial
tokens are concatenated without intervening full stops to generate
the initials component.
For an O/R address which follows the above restrictions, a string is
derived in the natural manner. In this case, the mapping will be
reversible.
For example:
GivenName = "Marshall"
Surname = "Rose"
Maps with "Marshall.Rose"
Initials = "MT"
Surname = "Rose"
Maps with "M.T.Rose"
GivenName = "Marshall"
Initials = "MT"
Surname = "Rose"
Maps with "Marshall.M.T.Rose"
Note that X.400 suggest that Initials is used to encode ALL initials.
Therefore, the defined encoding is "natural" when either GivenName or
Initials, but not both, are present. The case where both are present
can be encoded, but this appears to be contrived!
4.2.2. Standard Encoding of MTS.ORAddress
Given this structure, we can specify a BNF representation of an O/R
Address.
std-or-address = 1*( "/" attribute "=" value ) "/"
attribute = standard-type
/ "RFC-822"
/ registered-dd-type
/ dd-key "." std-printablestring
standard-type = key-string
registered-dd-type
= key-string
dd-key = key-string
value = std-printablestring
std-printablestring
= *( std-char / std-pair )
std-char = <"{", "}", "*", and any ps-char
except "/" and "=">
std-pair = "$" ps-char
The standard-type is any key defined in the table in Section 4.2,
except PN, and DD. The BNF leads to a set of attribute/value pairs.
The value is interpreted according to the EBNF encoding defined in
the table.
If the standard-type is PN, the value is interpreted according to
EBNF.encoded-pn, and the components of MTS.PersonalName and/or
MTS.TeletexPersonalName derived accordingly.
If dd-key is the recognised Domain Defined string (DD), then the type
and value are interpreted according to the syntax implied from the
encoding, and aligned to either the teletex or printable string form.
Key and value shall have the same encoding.
If value is "RFC-822", then the (printable string) Domain Defined
Type of "RFC-822" is assumed. This is an optimised encoding of the
domain defined type defined by this specification.
The matching of all keywords shall be done in a case-independent
manner.
EBNF.std-or-address uses the characters "/" and "=" as delimiters.
Domain Defined Attributes and any value may contain these characters.
A quoting mechanism, using the non-printable string "$" is used to
allow these characters to be represented.
If the value is registered-dd-type, and the value is registered at
the Internet Assigned Numbers Authority (IANA) as an accepted Domain
Defined Attribute type, then the value shall be interpreted
accordingly. This restriction maximises the syntax checking which
can be done at a gateway.
4.3. EBNF.822-address <-> MTS.ORAddress
Ideally, the mapping specified would be entirely symmetrical and
global, to enable addresses to be referred to transparently in the
remote system, with the choice of gateway being left to the Message
Transfer Service. There are two fundamental reasons why this is not
possible:
1. The syntaxes are sufficiently different to make this
awkward.
2. In the general case, there would not be the necessary
administrative co-operation between the X.400 and RFC822
worlds, which would be needed for this to work.
Therefore, an asymmetrical mapping is defined, which can be
symmetrical where there is appropriate administrative control.
4.3.1. X.400 encoded in RFC822
The std-or-address syntax is used to encode O/R Address information
in the 822.local-part of EBNF.822-address. In some cases, further
O/R Address information is associated with the 822.domain component.
This cannot be used in the general case, due to character set
problems, and to the variants of X.400 O/R Addresses which use
different attribute types. The only way to encode the full
PrintableString character set in a domain is by use of the
822.domain-ref syntax (i.e. 822.atom). This is likely to cause
problems on many systems. The effective character set of domains is
in practice reduced from the RFC822 set, by restrictions imposed by
domain conventions and policy, and by restrictions in RFC821.
A generic 822.address consists of a 822.local-part and a sequence of
822.domains (e.g., <@domain1,@domain2:user@domain3>). All except the
822.domain associated with the 822.local-part (domain3 in this case)
are considered to specify routing within the RFC822 world, and will
not be interpreted by the gateway (although they may have identified
the gateway from within the RFC822 world).
The 822.domain associated with the 822.local-part identifies the
gateway from within the RFC822 world. This final 822.domain may be
used to determine some number of O/R Address attributes, where this
does not conflict with the first role. RFC822 routing to gateways
will usually be set up to facilitate the 822.domain being used for
both purposes. The following O/R Address attributes are considered
as a hierarchy, and may be specified by the domain. They are (in
order of hierarchy):
Country, ADMD, PRMD, Organisation, Organisational Unit
There may be multiple Organisational Units.
A global mapping is defined between domain specifications, and some
set of attributes. This association proceeds hierarchically. For
example, if a domain implies ADMD, it also implies country.
Subdomains under this are associated according to the O/R Address
hierarchy. For example:
=> "AC.UK" might be associated with
C="GB", ADMD="GOLD 400", PRMD="UK.AC"
then domain "R-D.Salford.AC.UK" maps with
C="GB", ADMD="GOLD 400", PRMD="UK.AC", O="Salford", OU="R-D"
There are three basic reasons why a domain/attribute mapping might be
maintained, as opposed to using simply subdomains:
1. As a shorthand to avoid redundant X.400 information. In
particular, there will often be only one ADMD per country,
and so it does not need to be given explicitly.
2. To deal with cases where attribute values do not fit the
syntax:
domain-syntax = alphanum [ *alphanumhyphen alphanum ]
alphanum = <ALPHA or DIGIT>
alphanumhyphen = <ALPHA or DIGIT or HYPHEN>
Although RFC822 allows for a more general syntax, this
restricted syntax is chosen as it is the one chosen by the
various domain service administrations.
3. To deal with missing elements in the hierarchy. A domain
may be associated with an omitted attribute in conjunction
with several present ones. When performing the algorithmic
insertion of components lower in the hierarchy, the omitted
value shall be skipped. For example, if "HNE.EGM" is
associated with "C=TC", "ADMD=ECQ", "PRMD=HNE", and omitted
organisation, then "ZI.HNE.EGM" is mapped with "C=TC",
"ADMD=ECQ", "PRMD=HNE", "OU=ZI". Attributes may have null
values, and this is treated separately from omitted
attributes (whilst it would be bad practice to treat these
two cases differently, they must be allowed for).
This set of mappings needs be known by the gateways relaying between
the RFC822 world, and the O/R Address space associated with the
mapping in question. There needs to be a single global definition of
this set of mappings. A mapping implies an adminstrative equivalence
between the two parts of the namespaces which are mapped together.
To correctly route in all cases, it is necessary for all gateways to
know the mapping. To facilitate distribution of a global set of
mappings, a format for the exchange of this information is defined in
Appendix F.
The remaining attributes are encoded on the LHS, using the EBNF.std-
or-address syntax. For example:
/I=J/S=Linnimouth/GQ=5/@Marketing.Widget.COM
encodes the MTS.ORAddress consisting of:
MTS.CountryName = "TC"
MTS.AdministrationDomainName = "BTT"
MTS.OrganizationName = "Widget"
MTS.OrganizationalUnitNames.value = "Marketing"
MTS.PersonalName.surname = "Linnimouth"
MTS.PersonalName.initials = "J"
MTS.PersonalName.generation-qualifier = "5"
The first three attributes are determined by the domain Widget.COM.
Then, the first element of OrganizationalUnitNames is determined
systematically, and the remaining attributes are encoded on the LHS.
In an extreme case, all of the attributes will be on the LHS. As the
domain cannot be null, the RHS will simply be a domain indicating the
gateway.
The RHS (domain) encoding is designed to deal cleanly with common
addresses, and so the amount of information on the RHS is maximised.
In particular, it covers the Mnemonic O/R Address using a 1984
compatible encoding. This is seen as the dominant form of O/R
Address. Use of other forms of O/R Address, and teletex encoded
attributes will require an LHS encoding.
There is a further mechanism to simplify the encoding of common
cases, where the only attributes to be encoded on the LHS is a (non-
Teletex) Personal Name attributes which comply with the restrictions
of 4.2.1. To achieve this, the 822.local-part shall be encoded as
EBNF.encoded-pn. In the previous example, if the GenerationQualifier
was not present in the previous example O/R Address, it would map
with the RFC822 address: J.Linnimouth@Marketing.Widget.COM.
From the standpoint of the RFC822 Message Transfer System, the
domain specification is simply used to route the message in the
standard manner. The standard domain mechanisms are used to select
appropriate gateways for the corresponding O/R Address space. In
most cases, this will be done by registering the higher levels, and
assuming that the gateway can handle the lower levels.
4.3.2. RFC822 encoded in X.400
In some cases, the encoding defined above may be reversed, to give a
"natural" encoding of genuine RFC822 addresses. This depends
largely on the allocation of appropriate management domains.
The general case is mapped by use of domain defined attributes. A
Domain defined type "RFC-822" is defined. The associated attribute
value is an ASCII string encoded according to Section 3.3.3 of this
specification. The interpretation of the ASCII string depends on the
context of the gateway.
1. In the context of RFC822, and RFC920
[Crocker82a,Postel84a], the string can be used directly.
2. In the context of the JNT Mail protocol, and the NRS
[Kille84a,Larmouth83a], the string shall be interpreted
according to Mailgroup Note 15 [Kille84b].
3. In the context of UUCP based systems, the string shall be
interpreted as defined in [Horton86a].
Other O/R Address attributes will be used to identify a context in
which the O/R Address will be interpreted. This might be a
Management Domain, or some part of a Management Domain which
identifies a gateway MTA. For example:
C = "GB"
ADMD = "GOLD 400"
PRMD = "UK.AC"
O = "UCL"
OU = "CS"
"RFC-822" = "Jimmy(a)WIDGET-LABS.CO.UK"
OR
C = "TC"
ADMD = "Wizz.mail"
PRMD = "42"
"rfc-822" = "postel(a)venera.isi.edu"
Note in each case the PrintableString encoding of "@" as "(a)". In
the second example, the "RFC-822" domain defined attribute is
interpreted everywhere within the (Private) Management Domain. In
the first example, further attributes are needed within the
Management Domain to identify a gateway. Thus, this scheme can be
used with varying levels of Management Domain co-operation.
There is a limit of 128 characters in the length of value of a domain
defined attribute, and an O/R Address can have a maxmimum of four
domain defined attributes. Where the printable string generated from
the RFC822 address exceeeds this value, additional domain defined
attributes are used to enable up to 512 characters to be encoded.
These attributes shall be filled completely before the next one is
started. The DDA keywords are: RFC822C1; RFC822C2; RFC822C3.
Longer addresses cannot be encoded.
There is, analagous with 4.3.1, a means to associate parts of the O/R
Address hierarchy with domains. There is an analogous global
mapping, which in most cases will be the inverse of the domain to O/R
address mapping. The mapping is maintained separately, as there may
be differences (e.g., two alternate domain names map to the same set
of O/R address components).
4.3.3. Component Ordering
In most cases, ordering of O/R Address components is not significant
for the mappings specified. However, Organisational Units (printable
string and teletex forms) and Domain Defined Attributes are specified
as SEQUENCE in MTS.ORAddress, and so their order may be significant.
This specification needs to take account of this:
1. To allow consistent mapping into the domain hierarchy
2. To ensure preservation of order over multiple mappings.
There are three places where an order is specified:
1. The text encoding (std-or-address) of MTS.ORAddress as used
in the local-part of an RFC822 address. An order is needed
for those components which may have multiple values
(Organisational Unit, and Domain Defined Attributes). When
generating an 822.std-or-address, components of a given type
shall be in hierarchical order with the most significant
component on the RHS. If there is an Organisation
Attribute, it shall be to the right of any Organisational
Unit attributes. These requirements are for the following
reasons:
- Alignment to the hierarchy of other components in RFC
822 addresses (thus, Organisational Units will appear
in the same order, whether encoded on the RHS or LHS).
Note the differences of JNT Mail as described in
Appendix B.
- Backwards compatibility with RFC987/1026.
- To ensure that gateways generate consistent addresses.
This is both to help end users, and to generate
identical message ids.
Further, it is recommended that all other attributes are
generated according to this ordering, so that all attributes
so encoded follow a consistent hierarchy. When generating
822.msg-id, this order shall be followed.
2. For the Organisational Units (OU) in MTS.ORAddress, the
first OU in the SEQUENCE is the most significant, as
specified in X.400.
3. For the Domain Defined Attributes in MTS.ORAddress, the
First Domain Defined Attribute in the SEQUENCE is the most
significant.
Note that although this ordering is mandatory for this
mapping, there are NO implications on ordering significance
within X.400, where this is a Management Domain issue.
4.3.4. RFC822 -> X.400
There are two basic cases:
1. X.400 addresses encoded in RFC822. This will also include
RFC822 addresses which are given reversible encodings.
2. "Genuine" RFC822 addresses.
The mapping shall proceed as follows, by first assuming case 1).
STAGE I.
1. If the 822-address is not of the form:
local-part "@" domain
take the domain which will be routed on and apply step 2 of
stage 1 to derive (a possibly null) set of attributes. Then
go to stage II.
NOTE:It may be appropriate to reduce a source route address
to this form by removal of all bar the last domain. In
terms of the design intentions of RFC822, this would
be an incorrect action. However, in most real cases,
it will do the "right" thing and provide a better
service to the end user. This is a reflection on the
excessive and inappropriate use of source routing in
RFC822 based systems. Either approach, or the
intermediate approach of stripping only domain
references which reference the local gateway are
conformant to this specification.
2. Attempt to parse EBNF.domain as:
*( domain-syntax "." ) known-domain
Where EBNF.known-domain is the longest possible match in the
set of globally defined mappings (see Appendix F). If this
fails, and the EBNF.domain does not explicitly identify the
local gateway, go to stage II. If the domain explicitly
identifies the gateway, allocate no attributes. Otherwise,
allocate the attributes associated with EBNF.known-domain.
For each component, systematically allocate the attribute
implied by each EBNF.domain-syntax component in the order:
C, ADMD, PRMD, O, OU. Note that if the mapping used
identifies an "omitted attribute", then this attribute
should be omitted in the systematic allocation. If this new
component exceed an upper bound (ADMD: 16; PRMD: 16; O: 64;
OU: 32) or it would lead to more than four OUs, then go to
stage II with the attributes derived.
At this stage, a set of attributes has been derived, which
will give appropriate routing within X.400. If any of the
later steps of Stage I force use of Stage II, then these
attributes should be used in Stage II.
3. If the 822.local-part uses the 822.quoted-string encoding,
remove this quoting. If this unquoted 822.local-part has
leading space, trailing space, or two adjacent space go to
stage II.
4. If the unquoted 822.local-part contains any characters not
in PrintableString, go to stage II.
5. Parse the (unquoted) 822.local-part according to the EBNF
EBNF.std-or-address. Checking of upper bounds should not be
done at this point. If this parse fails, parse the local-
part according to the EBNF EBNF.encoded-pn. If this parse
fails, go to stage II. The result is a set of type/value
pairs. If the set of attributes leads to an address of any
form other than mnemonic form, then only these attributes
should be taken. If (for mnemonic form) the values generated
conflict with those derived in step 2 (e.g., a duplicated
country attribute), the domain is assumed to be a remote
gateway. In this case, take only the LHS derived
attributes, together with any RHS dericed attributes which
are more significant thant the most signicant attribute
which is duplicated (e.g., if there is a duplicate PRMD, but
no LHS derived ADMD and country, then the ADMD and country
should be taken from the RHS). therwise add LHS and RHS
derived attributes together.
6. Associate the EBNF.attribute-value syntax (determined from
the identified type) with each value, and check that it
conforms. If not, go to stage II.
7. Ensure that the set of attributes conforms both to the
MTS.ORAddress specification and to the restrictions on this
set given in X.400, and that no upper bounds are exceeded
for any attribute. If not go to stage II.
8. Build the O/R Address from this information.
STAGE II.
This will only be reached if the RFC822 EBNF.822-address is not a
valid X.400 encoding. This implies that the address must refer to a
recipient on an RFC822 system. Such addresses shall be encoded in
an X.400 O/R Address using a domain defined attribute.
1. Convert the EBNF.822-address to PrintableString, as
specified in Chapter 3.
2. Generate the "RFC-822" domain defined attribute from this
string.
3. Build the rest of the O/R Address in the manner described
below.
It may not be possible to encode the domain defined attribute due to
length restrictions. If the limit is exceeded by a mapping at the
MTS level, then the gateway shall reject the message in question. If
this occurs at the IPMS level, then the action will depend on the
policy being taken for IPMS encoding, which is discussed in Section
5.1.3.
If Stage I has identified a set of attributes, use these to build the
remainder of the address. The administrative equivalence of the
mappings will ensure correct routing throug X.400 to a gateway back
to RFC822.
If Stage I has not identified a set of attributes, the remainder of
the O/R address effectively identifies a source route to a gateway
from the X.400 side. There are three cases, which are handled
differently:
822-MTS Return Address
This shall be set up so that errors are returned through the
same gateway. Therefore, the O/R Address of the local
gateway shall be used.
IPMS Addresses
These are optimised for replying. In general, the message
may end up anywhere within the X.400 world, and so this
optimisation identifies a gateway appropriate for the RFC
822 address being converted. The 822.domain to which the
address would be routed is used to select an appropriate
gateway. A globally defined set of mappings is used, which
identifies (the O/R Address components of) appropriate
gateways for parts of the domain namespace. The longest
possible match on the 822.domain defines which gateway to
use. The table format for distribution of this information
is defined in Appendix F.
This global mapping is used for parts of the RFC822
namespace which do not have an administrative equivalence
with any part of the X.400 namespace, but for which it is
desirable to identify a preferred X.400 gateway in order to
optimise routing.
If no mapping is found for the 822.domain, a default value
(typically that of the local gateway) is used. It is never
appropriate to ignore the globally defined mappings. In
some cases, it may be appropriate to locally override the
globally defined mappings (e.g., to identify a gateway close
to a recipient of the message). This is likely to be where
the global mapping identifies a public gateway, and the
local gateway has an agreement with a private gateway which
it prefers to use.
822-MTS Recipient
As the RFC822 and X.400 worlds are fully connected, there
is no technical reason for this situation to occur. In some
cases, routing may be configured to connect two parts of the
RFC822 world using X.400. The information that this part
of the domain space should be routed by X.400 rather than
remaining within the RFC822 world will be configured
privately into the gateway in question. The O/R address
shall then be generated in the same manner as for an IPMS
address, using the globally defined mappings. It is to
support this case that the definition of the global domain
to gateway mapping is important, as the use of this mapping
will lead to a remote X.400 address, which can be routed by
X.400 routing procedures. The information in this mapping
shall not be used as a basis for deciding to convert a
message from RFC822 to X.400.
4.3.4.1. Heuristics for mapping RFC822 to X.400
RFC822 users will often use an LHS encoded address to identify an
X.400 recipient. Because the syntax is fairly complex, a number of
heuristics may be applied to facilitate this form of usage. A
gateway should take care not to be overly "clever" with heuristics,
as this may cause more confusion than a more mechanical approach.
The heuristics are as follows:
1. Ignore the omission of a trailing "/" in the std-or syntax.
2. If there is no ADMD component, and both country and PRMD are
present, the value of /ADMD= / (single space) is assumed.
3. Parse the unquoted local part according to the EBNF colon-
or-address. This may facilitate users used to this
delimiter.
colon-or-address = 1*(attribute "=" value ";" *(LWSP-char))
The remaining heuristic relates to ordering of address components.
The ordering of attributes may be inverted or mixed. For this
reason, the following heuristics may be applied:
4. If there is an Organisation attribute to the left of any Org
Unit attribute, assume that the hierarchy is inverted.
4.3.5. X.400 -> RFC822
There are two basic cases:
1. RFC822 addresses encoded in X.400.
2. "Genuine" X.400 addresses. This may include symmetrically
encoded RFC822 addresses.
When a MTS Recipient O/R Address is interpreted, gatewaying will be
selected if there is a single "RFC-822" domain defined attribute
present and the local gateway is identified by the remainder of the
O/R Address. In this case, use mapping A. For other O/R Addresses
which
1. Contain the special attribute.
AND
2. Identifies the local gateway or any other known gateway with
the other attributes.
use mapping A. In other cases, use mapping B.
NOTE:
A pragmatic approach would be to assume that any O/R
Address with the special domain defined attribute identifies
an RFC822 address. This will usually work correctly, but is
in principle not correct. Use of this approach is
conformant to this specification.
Mapping A
1. Map the domain defined attribute value to ASCII, as defined
in Chapter 3.
Mapping B
This is used for X.400 addresses which do not use the explicit RFC
822 encoding.
1. For all string encoded attributes, remove any leading or
trailing spaces, and replace adjacent spaces with a single
space.
The only attribute which is permitted to have zero length is
the ADMD. This should be mapped onto a single space.
These transformations are for lookup only. If an
EBNF.std-or-address mapping is used as in 4), then the
orginal values should be used.
2. Map numeric country codes to the two letter values.
3. Noting the hierarchy specified in 4.3.1 and including
omitted attributes, determine the maximum set of attributes
which have an associated domain specification in the
globally defined mapping. If no match is found, allocate
the domain as the domain specification of the local gateway,
and go to step 5.
Note: It might be appropriate to use a non-local domain.
This would be selected by a global mapping analagous to
the one described at the end of 4.3.4. This is not
done, primarily because use of RFC822 to connect X.400
systems is not expected to be significant.
In cases where the address refers to an X.400 UA, it is
important that the generated domain will correctly route to
a gateway. In general, this is achieved by carefully co-
ordinating RFC822 routing with the definition of the global
mappings, as there is no easy way for the gateway to make
this check. One rule that shall be used is that domains
with only one component will not route to a gateway. If the
generated domain does not route correctly, the address is
treated as if no match is found.
4. The mapping identified in 3) gives a domain, and an O/R
address prefix. Follow the hierarchy: C, ADMD, PRMD, O, OU.
For each successive component below the O/R address prefix,
which conforms to the syntax EBNF.domain-syntax (as defined
in 4.3.1), allocate the next subdomain. At least one
attribute of the X.400 address shall not be mapped onto
subdomain, as 822.local-part cannot be null. If there are
omitted attributes in the O/R address prefix, these will
have correctly and uniquely mapped to a domain component.
Where there is an attribute omitted below the prefix, all
attributes remaining in the O/R address shall be encoded on
the LHS. This is to ensure a reversible mapping. For
example, if the is an addres /S=XX/O=YY/ADMD=A/C=NN/ and a
mapping for /ADMD=A/C=NN/ is used, then /S=XX/O=YY/ is
encoded on the LHS.
5. If the address is not mnemonic form (form 1 variant 1),
then all of the attributes in the address should be encoded
on the LHS in EBNF.std-or-address syntax, as described
below.
For addresses of mnemonic form, if the remaining components
are personal-name components, conforming to the restrictions
of 4.2.1, then EBNF.encoded-pn is derived to form