4. Examples of PINT Requests and Responses
4.1. A request to a call center from an anonymous user to receive a
phone call.
C->S: INVITE sip:R2C@pint.mailorder.com SIP/2.0
Via: SIP/2.0/UDP 169.130.12.5
From: sip:anon-1827631872@chinet.net
To: sip:+1-201-456-7890@iron.org;user=phone
Call-ID: 19971205T234505.56.78@pager.com
CSeq: 4711 INVITE
Subject: Sale on Ironing Boards
Content-type: application/sdp
Content-Length: 174
v=0
o=- 2353687637 2353687637 IN IP4 128.3.4.5
s=R2C
i=Ironing Board Promotion
e=anon-1827631872@chinet.net
t=2353687637 0
m=audio 1 voice -
c=TN RFC2543 +1-201-406-4090
In this example, the context that is required to interpret the To:
address as a telephone number is not given explicitly; it is
implicitly known to the R2C@pint.mailorder.com server. But the
telephone of the person who wishes to receive the call is explicitly
identified as an internationally significant E.164 number that falls
within the North American numbering plan (because of the "+1" within
the c= line).
4.2. A request from a non anonymous customer (John Jones) to receive a
phone call from a particular sales agent (Mary James) concerning
the defective ironing board that was purchased
C->S: INVITE sip:marketing@pint.mailorder.com SIP/2.0
Via: SIP/2.0/UDP 169.130.12.5
From: sip:john.jones.3@chinet.net
To: sip:mary.james@mailorder.com
Call-ID: 19971205T234505.56.78@pager.com
CSeq: 4712 INVITE
Subject: Defective Ironing Board - want refund
Content-type: application/sdp
Content-Length: 150
v=0
o=- 2353687640 2353687640 IN IP4 128.3.4.5
s=marketing
e=john.jones.3@chinet.net
c= TN RFC2543 +1-201-406-4090
t=2353687640 0
m=audio 1 voice -
The To: line might include the Mary James's phone number instead of a
email-like address. An implementation that cannot accept email-like
URLs in the "To:" header must decline the request with a 606 Not
Acceptable. Note that the sending PINT client "knows" that the PINT
Gateway contacted with the "marketing@pint.mailorder.com" Request-URI
is capable of processing the client request as expected. (see 3.5.5.1
for a discussion on this).
Note also that such a telephone call service could be implemented on
the phone side with different details. For example, it might be that
first the agent's phone rings, and then the customer's phone rings,
or it might be that first the customer's phone rings and he hears
silly music until the agent comes on line. If necessary, such service
parameter details might be indicated in "a=" attribute lines within
the session description. The specification of such attribute lines
for service consistency is beyond the scope of the PINT 1.0
specifications.
4.3. A request from the same user to get a fax back on how to assemble
the Ironing Board
C->S: INVITE sip:faxback@pint.mailorder.com SIP/2.0
Via: SIP/2.0/UDP 169.130.12.5
From: sip:john.jones.3@chinet.net
To: sip:1-800-3292225@steam.edu;user=phone;phone-context=+1
Call-ID: 19971205T234505.66.79@chinet.net
CSeq: 4713 INVITE
Content-type: application/sdp
Content-Length: 218
v=0
o=- 2353687660 2353687660 IN IP4 128.3.4.5
s=faxback
e=john.jones.3@chinet.net
t=2353687660 0
m=application 1 fax URI
c=TN RFC2543 1-201-406-4091
a=fmtp:URI uri:http://localstore/Products/IroningBoards/2344.html
In this example, the fax to be sent is stored on some local server
(localstore), whose name may be only resolvable, or that may only be
reachable, from within the IP network on which the PINT server sits.
The phone number to be dialled is a "local phone number" as well.
There is no "phone-context" attribute, so the context (in this case,
for which nation the number is "nationally significant") must be
supplied by the faxback@pint.mailorder.com PINT server.
If the server that receives it does not understand the number, it
SHOULD decline the request and include a "Network Address Not
Understood" warning. Note that no "require" attribute was used here,
since it is very likely that the request can be serviced even by a
server that does not support the "require" attribute.
4.4. A request from same user to have that same information read out
over the phone
C->S: INVITE sip:faxback@pint.mailorder.com SIP/2.0
Via: SIP/2.0/UDP 169.130.12.5
From: sip:john.jones.3@chinet.net
To: sip:1-800-3292225@steam.edu;user=phone;phone-context=+1
Call-ID: 19971205T234505.66.79@chinet.net
CSeq: 4713 INVITE
Content-type: application/sdp
Content-Length: 220
v=0
o=- 2353687660 2353687660 IN IP4 128.3.4.5
s=faxback
e=john.jones.3@chinet.net
t=2353687660 0
m=application 1 voice URI
c=TN RFC2543 1-201-406-4090
a=fmtp:URI uri:http://localstore/Products/IroningBoards/2344.html
4.5. A request to send an included text page to a friend's pager.
In this example, the text to be paged out is included in the request.
C->S: INVITE sip:R2F@pint.pager.com SIP/2.0
Via: SIP/2.0/UDP 169.130.12.5
From: sip:scott.petrack@chinet.net
To: sip:R2F@pint.pager.com
Call-ID: 19974505.66.79@chinet.net
CSeq: 4714 INVITE
The image is available as tif or as gif. The tif is the preferred
format. Note that the http server where the pictures reside is local,
and the PINT server is also local (because it can resolve machine
name "petrack")
4.7. A request to read out over the phone two pieces of content in
sequence.
First some included text is read out by text-to-speech. Then some
text that is stored at some URI on the internet is read out.
C->S: INVITE sip:R2HC@pint.acme.com SIP/2.0
Via: SIP/2.0/UDP 169.130.12.5
From: sip:scott.petrack@chinet.net
To: sip:R2HC@pint.acme.com
Call-ID: 19974505.66.79@chinet.net
CSeq: 4716 INVITE
Content-Type: multipart/related; boundary=next
--next
Content-Type: application/sdp
Content-Length: 316
v=0
o=- 2353687720 2353687720 IN IP4 128.3.4.5
s=R2HC
e=scott.petrack@chinet.net
c= TN RFC2543 +1-201-406-4091
t=2353687720 0
m=text 1 voice plain
a=fmtp:plain spr:2@53655768
m=text 1 voice plain
a=fmtp:plain uri:http://www.your.com/texts/stuff.doc
--next
Content-Type: text/plain
Content-ID: 2@53655768
Content-Length: 172
Hello!! I am about to read out to you the document you
requested, "uri:http://www.your.com/texts/stuff.doc".
We hope you like acme.com's new speech synthesis server.
--next--
4.10. Sending a set of information in response to an enquiry
INVITE sip:R2FB@pint.bt.co.uk SIP/2.0
Via: SIP/2.0/UDP 169.130.12.5
To: sip:0345-12347-01@pint.bt.co.uk;user=phone;phone-context=+44
From: sip:colin.masterton@sales.hh.bt.co.uk
Call-ID: 19981205T234505.56.78@sales.hh.bt.co.uk
CSeq: 1147 INVITE
Subject: Price Info, as requested
Content-Type: multipart/related; boundary=next
--next
Content-type: application/sdp
Content-Length: 325
v=0
o=- 2353687780 2353687780 IN IP4 128.3.4.5
s=R2FB
i=Your documents
e=colin.masterton@sales.hh.bt.co.uk
t=2353687780 0
m=application 1 fax octet-stream
c=TN RFC2543 +44-1794-8331010
a=fmtp:octet-stream uri:http://www.bt.co.uk/imgs/pipr.gif opr:
spr:2@53655768
--next
Content-Type: text/plain
Content-ID: 2@53655768
Content-Length: 352
Dear Sir,
Thank you for your enquiry. I have checked availability in your
area, and we can provide service to your cottage. I enclose a
quote for the costs of installation, together with the ongoing
rental costs for the line. If you want to proceed with this,
please quote job reference isdn/hh/123.45.9901.
Yours Sincerely,
Colin Masterton
--next--
Note that the "implicit" faxback content is given by an EMPTY opaque
reference in the middle of the fmtp line in this example.
Note: in this case the opaque reference is a collection of data used
to convince the Executive System that the requester has the right to
get this information, rather than selecting the particular content
(the A party in the To: field of the SIP "wrapper" does that alone).
5. Security Considerations
5.1. Basic Principles for PINT Use
A PINT Gateway, and the Executive System(s) with which that Gateway
is associated, exist to provide service to PINT Requestors. The aim
of the PINT protocol is to pass requests from those users on to a
PINT Gateway so an associated Executive System can service those
requests.
5.1.1. Responsibility for service requests
The facility of making a GSTN-based call to numbers specified in the
PINT request, however, comes with some risks. The request can specify
an incorrect telephone of fax number. It is also possible that the
Requestor has purposely entered the telephone number of an innocent
third party. Finally, the request may have been intercepted on its
way through any intervening PINT or SIP infrastructure, and the
request may have been altered.
In any of these cases, the result may be that a call is placed
incorrectly. Where there is intent or negligence, this may be
construed as harassment of the person incorrectly receiving the call.
Whilst the regulatory framework for misuse of Internet connections
differs throughout the world and is not always mature, the rules
under which GSTN calls are made are much more settled. Someone may be
liable for mistaken or incorrect calls.
Understandably, the GSTN Operators would prefer that this someone is
not them, so they will need to ensure that any PINT Gateway and
Executive System combination does not generate incorrect calls
through some error in the Gateway or Executive system implementation
or GSTN-internal communications fault. Equally, it is important that
the Operator can show that they act only on requests that they have
good reason to believe are correct. This means that the Gateway must
not pass on requests unless it is sure that they have not been
corrupted in transit from the Requestor.
If a request can be shown to have come from a particular Requestor
and to have been acted on in good faith by the PINT service provider,
then responsibility for making requests may well fall to the
Requestor rather than the Operator who executed these requests.
Finally, it may be important for the PINT service provider to be able
to show that they act only on requests for which they have some
degree of assurance of origin. In many jurisdictions, it is a
requirement on GSTN Operators that they place calls only when they
can, if required, identify the parties to the call (such as when
required to carry out a Malicious Call Trace). It is at least likely
that the provider of PINT services will have a similar responsibility
placed on them.
It follows that the PINT service provider may require that the
identity of the Requestor be confirmed. If such confirmation is not
available, then they may be forced (or choose) not to provide
service. This identification may require personal authentication of
the Requesting User.
5.1.2. Authority to make requests
Where GSTN resources are used to provide a PINT service, it is at
least possible that someone will have to pay for it. This person may
not be the Requestor, as, for example, in the case of existing GSTN
split-charging services like free phone in which the recipient of a
call rather than the originator is responsible for the call cost.
This is not, of course, the only possibility; for example, PINT
service may be provided on a subscription basis, and there are a
number of other models. However, whichever model is chosen, there may
be a requirement that the authority of a Requestor to make a PINT
request is confirmed.
If such confirmation is not available, then, again, the PINT Gateway
and associated Executive System may choose not to provide service.
5.1.3. Privacy
Even if the identity of the Requesting User and the Authority under
which they make their request is known, there remains the possibility
that the request is either corrupted, maliciously altered, or even
replaced whilst in transit between the Requestor and the PINT
Gateway.
Similarly, information on the Authority under which a request is made
may well be carried within that request. This can be sensitive
information, as an eavesdropper might steal this and use it within
their own requests. Such authority SHOULD be treated as if it were
financial information (such as a credit card number or PIN).
The data authorizing a Requesting User to make a PINT request should
be known only to them and the service provider. However, this
information may be in a form that does not match the schemes normally
used within the Internet. For example, X.509 certificates[14] are
commonly used for secured transactions on the Internet both in the IP
Security Architecture[12] and in the TLS protocol[13], but the GSTN
provider may only store an account code and PIN (i.e. a fixed string
of numbers).
A Requesting User has a reasonable expectation that their requests
for service are confidential. For some PINT services, no content is
carried over the Internet; however, the telephone or fax numbers of
the parties to a resulting service calls may be considered sensitive.
As a result, it is likely that the Requestor (and their PINT service
provider) will require that any request that is sent across the
Internet be protected against eavesdroppers; in short, the requests
SHOULD to be encrypted.
5.1.4. Privacy Implications of SUBSCRIBE/NOTIFY
Some special considerations relate to monitoring sessions using the
SUBSCRIBE and NOTIFY messages. The SUBSCRIBE message that is used to
register an interest in the disposition of a PINT service transaction
uses the original Session Description carried in the related INVITE
message. This current specification does not restrict the source of
such a SUBSCRIBE message, so it is possible for an eavesdropper to
capture an unprotected session description and use this in a
subsequent SUBSCRIBE request. In this way it is possible to find out
details on that transaction that may well be considered sensitive.
The initial solution to this risk is to recommend that a session
description that may be used within a subsequent SUBSCRIBE message
SHOULD be protected.
However, there is a further risk; if the origin-field used is
"guessable" then it might be possible for an attacker to reconstruct
the session description and use this reconstruction within a
SUBSCRIBE message.
SDP (see section 6 of [2], "o=" field) does not specify the mechanism
used to generate the sess-id field, and suggests that a method based
on timestamps produced by Network Time Protocol [16] can be used.
This is sufficient to guarantee uniqueness, but may allow the value
to be guessed, particularly if other unprotected requests from the
same originator are available.
Thus, to ensure that the session identifier is not guessable the
techniques described in section 6.3 of [17] can be used when
generating the origin-field for a session description to be used
inside a PINT INVITE message. If all requests from (and responses to)
a particular PINT requesting entity are protected, then this is not
needed. Where such a situation is not assured, AND where session
monitoring is supported, then a method by which an origin-field
within a session description is not guessable SHOULD be used.
5.2. Registration Procedures
Any number of PINT Gateways may register to provide the same service;
this is indicated by the Gateways specifying the same "userinfo" part
in the To: header field of the REGISTER request. Whilst such
ambiguity would be unlikely to occur with the scenarios covered by
"core" SIP, it is very likely for PINT; there could be any number of
service providers all willing to support a "Request-To-Fax" service,
for example.
Unless a request specifies the Gateway name explicitly, an
intervening Proxy that acts on a registration database to which
several Gateways have all registered is in a position to select from
the registrands using whatever algorithm it chooses; in principle,
any Gateway that has registered as "R2F" would be appropriate.
However, this opens up an avenue for attack, and this is one in which
a "rogue" Gateway operator stands to make a significant gain. The
standard SIP procedure for releasing a registration is to send a
REGISTER request with a Contact field having a wildcard value and an
expires parameter with a value of 0. It is important that a PINT
Registrar uses authentication of the Registrand, as otherwise one
PINT service provider would be able to "spoof" another and remove
their registration. As this would stop the Proxy passing any requests
to that provider, this would both increase requests being sent to the
rogue and stop requests going to the victim.
Another variant on this attack would be to register a Gateway using a
name that has been registered by another provider; thus a rogue
Operator might register its Gateway as "R2C@pint.att.com", thereby
hijacking requests.
The solution is the same; all registrations by PINT Gateways MUST be
authenticated; this includes both new or apparent replacement
registrations, and any cancellation of current registrations. This
recommendation is also made in the SIP specification, but for the
correct operation of PINT, it is very important indeed.
5.3. Security mechanisms and implications on PINT service
PINT is a set of extensions to SIP[1] and SDP[2], and will use the
security procedures described in SIP. There are several implications
of this, and these are covered here.
For several of the PINT services, the To: header field of SIP is used
to identify one of the parties to the resulting service call. The
PINT Request-To-Call service is an example. As mentioned in the SIP
specification, this field is used to route SIP messages through an
infrastructure of Redirect and Proxy server between the corresponding
User Agent Servers, and so cannot be encrypted. This means that,
although the majority of personal or sensitive data can be protected
whilst in transit, the telephone (or fax) number of one of the
parties to a PINT service call cannot, and will be "visible" to any
interception. For the PINT milestone services this may be acceptable,
since the caller named in the To: service is typically a "well known"
provider address, such as a Call Center.
Another aspect of this is that, even if the Requesting User does not
consider the telephone or fax numbers of the parties to a PINT
service to be private, those parties might. Where PINT servers have
reason to believe this might be the case they SHOULD encrypt the
request, even if the Requestor has not done so. This could happen,
for example, if a Requesting User within a company placed a PINT
request and this was carried via the company's Intranet to their
Proxy/firewall and thence over the Internet to a PINT Gateway at
another location.
If a request carries data that can be reused by an eavesdropper
either to "spoof" the Requestor or to obtain PINT service by
inserting the Requestor's authorization token into an eavesdropper's
request, then this data MUST be protected. This is particularly
important if the authorization token consists of static text (such as
an account code and/or PIN).
One approach is to encrypt the whole of the request, using the
methods described in the SIP specification. As an alternative, it may
be acceptable for the authorization token to be held as an opaque
reference (see section 3.4.2.3 and examples 4.11 and 4.12), using
some proprietary scheme agreed between the Requestor and the PINT
service provider, as long as this is resistant to interception and
re-use. Also, it may be that the authorization token cannot be used
outside of a request cryptographically signed by the Requestor; if so
then this requirement can be relaxed, as in this case the token
cannot be re-used by another. However, unless both the Requestor and
the Gateway are assured that this is the case, any authorization
token MUST be treated as sensitive, and so MUST be encrypted.
A PINT request may contain data within the SDP message body that can
be used more efficiently to route that request. For example, it may
be that one Gateway and Executive System combination cannot handle a
request that specifies one of the parties as a pager, whilst another
can. Both gateways may have registered with a PINT/SIP Registrar, and
this information may be available to intervening PINT/SIP Proxies.
However, if the message body is encrypted, then the request cannot be
decoded at the Proxy server, and so Gateway selection based on
contained information cannot be made there.
The result is that the Proxy may deliver the request to a Gateway
that cannot handle it; the implication is that a PINT/SIP Proxy
SHOULD consider its choice for the appropriate Gateway subject to
correction, and, on receiving a 501 or 415 rejection from the first
gateway chosen, try another. In this way, the request will succeed if
at all possible, even though it may be delayed (and tie up resources
in the inappropriate Gateways).
This opens up an interesting avenue for Denial Of Service; sending a
valid request that appears to be suitable for a number of different
Gateways, and simply occupying those Gateways in decrypting a message
requesting a service they cannot provide. As mentioned in section
3.5.5.1, the choice of service name to be passed in the userinfo
portion of the SIP Request-URI is flexible, and it is RECOMMENDED
that names be chosen that allow a Proxy to select an appropriate
Gateway without having to examine the SDP body part. Thus, in the
example given here, the service might be called "Request-To-Page" or
"R2P" rather than the more general use of "R2F", if there is a
possibility of the SDP body part being protected during transit.
A variation on this attack is to provide a request that is
syntactically invalid but that, due to the encryption, cannot be
detected without expending resources in decoding it. The effects of
this form of attack can be minimised in the same way as for any SIP
Invitation; the Proxy should detect the 400 rejection returned from
the initial Gateway, and not pass the request onwards to another.
Finally, note that the Requesting User may not have a prior
relationship with a PINT Gateway, whilst still having a prior
relationship with the Operator of the Executive System that fulfills
their request. Thus there may be two levels of authentication and
authorization; one carried out using the techniques described in the
SIP specification (for use between the Requestor and the Gateway),
with another being used between the Requesting User or the Requestor
and the Executive System.
For example, the Requesting User may have an account with the PINT
service provider. That provider might require that requests include
this identity before they will be convinced to provide service. In
addition, to counter attacks on the request whilst it is in transit
across the Internet, the Gateway may require a separate X.509-based
certification of the request. These are two separate procedures, and
data needed for the former would normally be expected to be held in
opaque references inside the SDP body part of the request.
The detailed operation of this mechanism is, by definition, outside
the scope of an Internet Protocol, and so must be considered a
private matter. However, one approach to indicating to the Requestor
that such "second level" authentication or authorization is required
by their Service Provider would be to ask for this inside the textual
description carried with a 401 response returned from the PINT
Gateway.
5.4. Summary of Security Implications
From the above discussion, PINT always carries data items that are
sensitive, and there may be financial considerations as well as the
more normal privacy concerns. As a result, the transactions MUST be
protected from interception, modification and replay in transit.
PINT is based on SIP and SDP, and can use the security procedures
outlined in [1] (sections 13 and 15). However, in the case of PINT,
the SIP recommendation that requests and responses MAY be protected
is not enough. PINT messages MUST be protected, so PINT
Implementations MUST support SIP Security (as described in [1],
sections 13 & 15), and be capable of handling such received messages.
In some configurations, PINT Clients, Servers, and Gateways can be
sure that they operate using the services of network level security
[13], transport layer security [12], or physical security for all
communications between them. In these cases messages MAY be exchanged
without SIP security, since all traffic is protected already. Clients
and servers SHOULD support manual configuration to use such lower
layer security facilities.
When using network layer security [13], the Security Policy Database
MUST be configured to provide appropriate protection to PINT traffic.
When using TLS, a port configured MUST NOT also be configured for
non-TLS traffic. When TLS is used, basic authentication MUST be
supported, and client-side certificates MAY be supported.
Authentication of the Client making the request is required, however,
so if this is not provided by the underlying mechanism used, then it
MUST be included within the PINT messages using SIP authentication
techniques. In contrast with SIP, PINT requests are often sent to
parties with which a prior communications relationship exists (such
as a Telephone Carrier). In this case, there may be a shared secret
between the client and the PINT Gateway. Such PINT systems MAY use
authentication based on shared secrets, with HTTP "basic
authentication". When this is done, the message integrity and privacy
must be guaranteed by some lower layer mechanism.
There are implications on the operation of PINT here though. If a
PINT proxy or redirect server is used, then it must be able to
examine the contents of the IP datagrams carried. It follows that an
end-to-end approach using network-layer security between the PINT
Client and a PINT Gateway precludes the use of an intervening proxy;
communication between the Client and Gateway is carried via a tunnel
to which any intervening entity cannot gain access, even if the IP
datagrams are carried via this node. Conversely, if a "hop-by-hop"
approach is used, then any intervening PINT proxies (or redirect
servers) are, by implication, trusted entities.
However, if there is any doubt that there is an underlying network or
transport layer security association in place, then the players in a
PINT protocol exchange MUST use encryption and authentication
techniques within the protocol itself. The techniques described in
section 15 of RFC2543 MUST be used, unless there is an alternative
protection scheme that is agreed between the parties. In either case,
the content of any message body (or bodies) carried within a PINT
request or response MUST be protected; this has implications on the
options for routing requests via Proxies (see 5.3).
Using SIP techniques for protection, the Request-URI and To: fields
headers within PINT requests cannot be protected. In the baseline
PINT services these fields may contain sensitive information. This is
a consideration, and if these data ARE considered sensitive, then
this will preclude the sole use of SIP techniques; in such a
situation, transport [12] or network layer [13] protection mechanisms
MUST be used.
As a final point, this choice will in turn have an influence on the
choice of transport layer protocol that can be used; if a TLS
association is available between two nodes, then TCP will have to be
used. This is different from the default behaviour of SIP (try UDP,
then try TCP if that fails).
6. Deployment considerations and the Relationship PINT to I.N.
(Informative)
6.1. Web Front End to PINT Infrastructure
It is possible that some other protocol may be used to communicate a
Requesting User's requirements. Due to the high numbers of available
Web Browsers and servers it seems likely that some PINT systems will
use HTML/HTTP as a "front end". In this scenario, HTTP will be used
over a connection from the Requesting User's Web Browser (WC) to an
Intermediate Web Server (WS). This will be closely associated with a
PINT Client (using some unspecified mechanism to transfer the data
from the Web Server to the PINT Client). The PINT Client will
represent the Requesting User to the PINT Gateway, and thus to the
Executive System that carries out the required action.
[WC]------[WS]
[PC]
\
\
[PG]
[XS]
Figure 2: Basic "Web-fronted" Configuration
6.2. Redirects to Multiple Gateways
It is quite possible that a given PINT Gateway is associated with an
Executive System (or systems) that can connect to the GSTN at
different places. Equally, if there is a chain of PINT Servers, then
each of these intermediate or proxy servers (PP) may be able to route
PINT requests to Executive Systems that connect at specific points to
the GSTN. The result of this is that there may be more than one PINT
Gateway or Executive System that can deal with a given request. The
mechanisms by which the choice on where to deliver a request are
outside the scope of this document.
[WC]------[WS] [WC]------[WS]
[PC] [PC]
\ \
\ \
[PG] [PP]
.........[XS]......... / \
: : / \
[PG] [PG]
[XS] [XS]
Figure 3: Multiple Access Configurations
However, there do seem to be two approaches. Either a Server that
acts as a proxy or redirect will select the appropriate Gateway
itself and will cause the request to be sent on accordingly, or a
list of possible Locations will be returned to the Requesting User
from which they can select their choice.
In SIP, the implication is that, if a proxy cannot resolve to a
single unique match for a request destination, then a response
containing a list of the choices should be returned to the Requesting
User for selection. This is not too likely a scenario within the
normal use of SIP.
However, within PINT, such ambiguity may be quite common; it implies
that there are a number of possible providers of a given service.
6.3. Competing PINT Gateways REGISTERing to offer the same service
With PINT, the registration is not for an individual but instead for
a service that can be handled by a service provider. Thus, one can
envisage a registration by the PINT Server of the domain telcoA.com
of its ability to support the service R2C as "R2C@telcoA.com", sent
to an intermediary server that acts as registrar for the
"broker.telcos.com" domain from "R2C@pint.telcoA.com" as follows:
REGISTER sip:registrar@broker.telcos.com SIP/2.0
To: sip:R2C@pint.telcoA.com
From: sip:R2C@pint.telcoA.com
...
This is the standard SIP registration service.
However, what happens if there are a number of different Service
Providers, all of whom support the "R2C" service? Suppose there is a
PINT system at domain "broker.com". PINT clients requesting a
Request-to-Call service from broker.com might be very willing to be
redirected or proxied to any one of the various service providers
that had previously registered with the registrar. PINT servers might
also be interested in providing service for requests that did not
specify the service provider explicitly, as well as those requests
that were directed "at them".
To enable such service, PINT servers would REGISTER at the broker
PINT server registrations of the form:
REGISTER sip:registrar@broker.com SIP/2.0
To: sip:R2C@broker.com
From: sip:R2C@pint.telcoA.com
When several such REGISTER messages appear at the registrar, each
differing only in the URL in the From: line, the registrar has many
possibilities, e.g.:
(i) it overwrites the prior registration for "R2C@broker.telcos.com"
when the next comes in;
(ii) it rejects the subsequent registration for
"R2C@broker.telcos.com";
(iii) it maintains all such registrations.
In this last case, on receiving an Invitation for the "general"
service, either:
(iii.1) it passes on the invitation to all registered service
providers, returning a collated response with all
acceptances, using multiple Location: headers,
or
(iii.2) it silently selects one of the registrations (using, for
example, a "round robin" approach) and routes the Invitation
and response onwards without further comment.
As an alternative to all of the above approaches, it:
(iv) may choose to not allow registrations for the "general" service,
rejecting all such REGISTER requests.
The algorithm by which such a choice is made will be implementation-
dependent, and is outside the scope of PINT. Where a behaviour is to
be defined by requesting users, then some sort of call processing
language might be used to allow those clients, as a pre-service
operation, to download the behaviour they expect to the server making
such decisions. This, however, is a topic for other protocols, not
for PINT.
6.4. Limitations on Available Information and Request Timing for
SUBSCRIBE
A reference configuration for PINT is that service requests are sent,
via a PINT Gateway, to an Executive System that fulfills the Service
Control Function (SCF) of an Intelligent Network (see [11]). The
success or failure of the resulting service call may be information
available to the SCF and so may potentially be made available to the
PINT Gateway. In terms of historical record of whether or not a
service succeeded, a large SCF may be dealing with a million call
attempts per hour. Given that volume of service transactions, there
are finite limits beyond which it cannot store service disposition
records; expecting to find out if a Fax was sent last month from a
busy SCF is unrealistic.
Other status changes, such as that on completion of a successful
service call, require the SCF to arrange monitoring of the service
call in a way that the service may not do normally, for performance
reasons. In most implementations, it is difficult efficiently to
interrupt a service to change it once it has begun execution, so it
may be necessary to have two different services; one that sets GSTN
resources to monitor service call termination, and one that doesn't.
It is unlikely to be possible to decide that monitoring is required
once the service has started.
These factors can have implications both on the information that is
potentially available at the PINT Gateway, and when a request to
register interest in the status of a PINT service can succeed. The
alternative to using a general SCF is to provide a dedicated Service
Node just for PINT services. As this node is involved in placing all
service calls, it is in a position to collect the information needed.
However, it may well still not be able to respond successfully to a
registration of interest in call state changes once a service logic
program instance is running.
Thus, although a Requesting User may register an interest in the
status of a service request, the PINT Gateway may not be in a
position to comply with that request. Although this does not affect
the protocol used between the Requestor and the PINT Gateway, it may
influence the response returned. To avoid the problem of changing
service logic once running, any registration of interest in status
changes should be made at or before the time at which the service
request is made.
Conversely, if a historical request is made on the disposition of a
service, this should be done within a short time after the service
has completed; the Executive System is unlikely to store the results
of service requests for long; these will have been processed as AMA
(Automatic Message Accounting) records quickly, after which the
Executive System has no reason to keep them, and so they may be
discarded.
Where the PINT Gateway and the Executive System are intimately
linked, the Gateway can respond to status subscription requests that
occur while a service is running. It may accept these requests and
simply not even try to query the Executive System until it has
information that a service has completed, merely returning the final
status. Thus the PINT Requestor may be in what it believes is a
monitoring state, whilst the PINT Gateway has not even informed the
Executive System that a request has been made. This will increase the
internal complexity of the PINT Gateway in that it will have a
complex set of interlocking state machines, but does mean that status
registration and indication CAN be provided in conjunction with an
I.N. system.
6.5. Parameters needed for invoking traditional GSTN Services within
PINT
This section describes how parameters needed to specify certain
traditional GSTN services can be carried within PINT requests.
6.5.1. Service Identifier
When a Requesting User asks for a service to be performed, he or she
will, of course, have to specify in some way which service. This can
be done in the URLs within the To: header and the Request-URI (see
section 3.5.5.1).
6.5.2. A and B parties
With the Request-to-Call service, they will also need to specify the
A and B parties they want to be engaged in the resulting service
call. The A party could identify, for example, the Call Center from
which they want a call back, whilst the B party is their telephone
number (i.e. who the Call Center agent is to call).
The Request-to-Fax and Request-to-Hear-Content services require the B
party to be specified (respectively the telephone number of the
destination Fax machine or the telephone to which spoken content is
to be delivered), but the A party is a Telephone Network based
resource (either a Fax or speech transcoder/sender), and is implicit;
the Requesting User does not (and cannot) specify it.
With the "Fax-Back" variant of the Request-to-Fax service, (i.e.
where the content to be delivered resides on the GSTN) they will also
have specify two parties. As before, the B party is the telephone
number of the fax machine to which they want a fax to be sent.
However, within this variant the A party identifies the "document
context" for the GSTN-based document store from which a particular
document is to be retrieved; the analogy here is to a GSTN user
dialling a particular telephone number and then entering the document
number to be returned using "touch tone" digits. The telephone number
they dial is that of the document store or A party, with the "touch
tone" digits selecting the document within that store.
6.5.3. Other Service Parameters
In terms of the extra parameters to the request, the services again
differ. The Request-to-Call service needs only the A and B parties.
Also it is convenient to assert that the resulting service call will
carry voice, as the Executive System within the destination GSTN may
be able to check that assertion against the A and B party numbers
specified and may treat the call differently.
With the Request-to-Fax and Request-to-Hear-Content services, the
source information to be transcoded is held on the Internet. That
means either that this information is carried along with the request
itself, or that a reference to the source of this information is
given.
In addition, it is convenient to assert that the service call will
carry fax or voice, and, where possible, to specify the format for
the source information.
The GSTN-based content or "Fax-Back" variant of the Request-to-Fax
service needs to specify the Document Store number and the Fax
machine number to which the information is to be delivered. It is
convenient to assert that the call will carry Fax data, as the
destination Executive System may be able to check that assertion
against the document store number and that of the destination Fax
machine.
In addition, the document number may also need to be sent. This
parameter is an opaque reference that is carried through the Internet
but has significance only within the GSTN. The document store number
and document number together uniquely specify the actual content to
be faxed.
6.5.4. Service Parameter Summary
The following table summarises the information needed in order to
specify fully the intent of a GSTN service request. Note that it
excludes any other parameters (such as authentication or
authorisation tokens, or Expires: or CallId: headers) that may be
used in a request.
Service ServiceID AParty BParty CallFmt Source SourceFmt
------- --------- ------ ------ ------- ------ -------
R2C x x x voice - -
R2F x - x fax URI/IL ISF/ILSF
R2FB x x x fax OR -
R2HC x - x voice URI/IL ISF/ILSF
In this table, "x" means that the parameter is required, whilst "-"
means that the parameter is not required.
The Services listed are Request-to-Call (R2C), Request-to-Fax (R2F),
the GSTN-based content or "Fax-back" Variant of Request-to-Fax
(R2FB), and Request-to-Hear-Content (R2HC).
The Call Format parameter values "voice" or "fax" indicate the kind
of service call that results.
The Source Indicator "URI/IL" implies that the information is either
an Internet source reference (a Universal Resource Identifier, or
URI) or is carried "in-line" with the message. The Source indicator
"OR" means that the value passed is an Opaque Reference that should
be carried along with the rest of the message but is to be
interpreted only within the destination (GSTN) context. As an
alternative, it could be given as a "local" reference with the "file"
style, or even using a partial reference with the "http" style.
However, the way in which such a reference is interpreted is a matter
for the receiving PINT Server and Executive System; it remains, in
effect, an opaque reference.
The Source Format value "ISF/ILSF" means that the format of the
source is specified either in terms of the URI or that it is carried
"in-line". Note that, for some data, the format either can be
detected by inspection or, if all else fails, can be assumed from the
URI (for example, by assuming that the file extension part of a URL
indicates the data type). For an opaque reference, the Source Format
is not available on the Internet, and so is not given.
6.6. Parameter Mapping to PINT Extensions
This section describes the way in which the parameters needed to
specify a GSTN service request fully might be carried within a "PINT
extended" message. There are other choices, and these are not
precluded. However, in order to ensure that the Requesting User
receives the service that they expect, it is necessary to have some
shared understanding of the parameters passed and the behaviour
expected of the PINT Server and its attendant Executive System.
The Service Identifier can be sent as the userinfo element of the
Request-URI. Thus, the first line of a PINT Invitation would be of
the form:
INVITE <serviceID>@<pint-server>.<domain> SIP/2.0
The A Party for the Request-to-Call and "Fax-back" variant of
Request-to-Fax service can be held in the "To:" header field. In this
case the "To:" header value will be different from the Request-URI.
In the services where the A party is not specified, the "To:" field
is free to repeat the value held in the Request-URI. This is the case
for Request-to-Fax and Request-to-Hear-Content services.
The B party is needed in all these milestone services, and can be
held in the enclosed SDP sub-part, as the value of the "c=" field.
The call format parameter can be held as part of the "m=" field
value. It maps to the "transport protocol" element as described in
section 3.4.2 of this document.
The source format specifier is held in the "m=", as a type and either
"-" or sub-type. The latter is normally required for all services
except Request-to-Call or "Faxback", where the "-" form may be used.
As shown earlier, the source format and source are not always
required when generating requests for services. However, the
inclusion in all requests of a source format specifier can make
parsing the request simpler and allows for other services to be
specified in the future, and so values are always given. The source
format parameter is covered in section 3.4.2 as the "media type"
element.
The source itself is identified by an "a=fmtp:" field value, where
needed. With the exception of the Request-to-Call service, all
invitations will normally include such a field. From the perspective
of the SDP extensions, it can be considered as qualifying the media
sub-type, as if to say, for example, "when I say jpeg, what I mean is
the following".
In summary, the parameters needed by the different services are
carried in fields as shown in the following table:
Service Svc Param PINT/SIP or SDP field used Example value
------- --------- -------------------------- -------------
R2C
ServiceID: <SIP Request-URI userinfo> R2C
AParty: <SIP To: field> sip:123@p.com
BParty: <SDP c= field> TN RFC2543 4567
CallFormat: <SDP transport protocol
sub-field of m= field> voice
SourceFmt: <SDP media type sub-field
of m= field> audio
(--- only "-" sub-type
sub-field value used) ---
Source: (--- No source specified) ---
R2F
ServiceID: <SIP Request-URI userinfo> R2F
AParty: (--- SIP To: field not used) sip:R2F@pint.xxx.net
BParty: <SDP c= field> TN RFCxxx +441213553
CallFormat: <SDP transport protocol
sub-field of m= field> fax
SourceFmt: <SDP media type sub-field
of m= field> image
<SDP media sub-type sub-field
of m= field> jpeg
Source: <SDP a=fmtp: field qualifying
preceding m= field> a=fmtp:jpeg<uri-ref>
R2FB
ServiceID: <SIP Request-URI userinfo> R2FB
AParty: <SIP To: field> sip:1-730-1234@p.com
BParty: <SDP c= field> TN RFCxxx +441213553
CallFormat: <SDP transport protocol
sub-field of m= field> fax
SourceFmt: <SDP media type sub-field
of m= field> image
<SDP media sub-type sub-field
of m= field> jpeg
Source: <SDP a=fmtp: field qualifying
preceding m= field> a=fmtp:jpeg opr:1234
R2HC
ServiceID: <SIP Request-URI userinfo> R2HC
AParty: (--- SIP To: field not used) sip:R2HC@pint.ita.il
BParty: <SDP c= field> TN RFCxxx +441213554
CallFormat: <SDP transport protocol
sub-field of m= field> voice
SourceFmt: <SDP media type sub-field
of m= field> text
<SDP media sub-type sub-field
of m= field> html
Source: <SDP a=fmtp: field qualifying
preceding m= field> a=fmtp:html<uri-ref>