One document matched: draft-ietf-shim6-multihome-shim-api-07.xml
<?xml version="1.0"?>
<?xml-stylesheet type='text/xsl' href='ref2629.xslt' ?>
<!DOCTYPE rfc SYSTEM "rfc2629.dtd">
<rfc category="info" docName="draft-ietf-shim6-multihome-shim-api-07"
ipr="full3978">
<?xml-stylesheet type='text/xsl' href='rfc2629.xslt' ?> <?rfc
toc="yes" ?> <?rfc symrefs="yes" ?> <?rfc sortrefs="yes" ?> <?rfc
iprnotified="no" ?> <?rfc strict="yes" ?>
<?rfc compact="yes"?>
<front>
<title abbrev="Multihoming Shim API">Socket Application Program
Interface (API) for Multihoming Shim</title>
<author initials='M' surname='Komu' fullname='Miika Komu'>
<organization abbrev="HIIT">Helsinki Institute for Information
Technology</organization>
<address>
<postal>
<street>Tammasaarenkatu 3</street>
<city>Helsinki</city>
<country>Finland</country>
</postal>
<phone>+358503841531</phone>
<facsimile>+35896949768</facsimile>
<email>miika@iki.fi</email>
<uri>http://www.hiit.fi/</uri>
</address>
</author>
<author initials='M' surname='Bagnulo' fullname='Marcelo Bagnulo'>
<organization abbrev="UC3M">Universidad Carlos III de
Madrid</organization>
<address>
<postal>
<street>Av. Universidad 30</street>
<city>Leganes</city>
<code>28911</code>
<country>SPAIN</country>
</postal>
<phone>+34 91 6248837</phone>
<email>marcelo@it.uc3m.es</email>
<uri>http://it.uc3m.es/marcelo</uri>
</address>
</author>
<author initials='K' surname='Slavov' fullname='Kristian Slavov'>
<organization abbrev="Ericsson">Ericsson Research
Nomadiclab</organization>
<address>
<postal>
<street>Hirsalantie 11</street>
<city>Jorvas</city>
<code>FI-02420</code>
<country>Finland</country>
</postal>
<phone>+358 9 299 3286</phone>
<email>kristian.slavov@ericsson.com</email>
</address>
</author>
<author initials='S' surname='Sugimoto' fullname='Shinta Sugimoto'
role="editor">
<organization abbrev="Ericsson"> Nippon Ericsson K.K.</organization>
<address>
<postal>
<street>Koraku Mori Building</street>
<street>1-4-14, Koraku, Bunkyo-ku</street>
<city>Tokyo</city>
<code>112-0004</code>
<country>Japan</country>
</postal>
<phone>+81 3 3830 2241</phone>
<email>shinta.sugimoto@ericsson.com</email>
</address>
</author>
<date month="November" day="3" year="2008"/>
<area>Internet</area>
<workgroup>SHIM6 Working Group</workgroup>
<keyword>SHIM6, HIP, identifier/locator split</keyword>
<abstract>
<t>This document specifies sockets API extensions for the
multihoming shim layer. The API aims to enable interactions
between applications and the multihoming shim layer for advanced
locator management, and access to information about failure
detection and path exploration.</t>
<t>This document is based on an assumption that a multihomed
host is equipped with a conceptual sub-layer (hereafter "shim")
inside the IP layer that maintains mappings between identifiers
and locators. Examples of the shim are SHIM6 and HIP.</t>
</abstract>
</front>
<middle>
<!--
================================================================
Introduction
================================================================
-->
<section title="Introduction" toc="include">
<t>HIP and SHIM6 have a commonality in their protocol design in
the sense that the roles of an IP address as an identifier and a
locator are clearly distinguished. Hereafter this design
principle is called "identifier/locator separation" in this
document. Both protocols aim to solve problems that are
specific to multihoming environment in an endhost centric
approach. In these protocols, a sub-layer within the IP layer
maintains mappings of identifiers and locators.</t>
<t>The shim layer is useful in a sense that the IP layer can
maintain the mapping of an identifier to the corresponding
locators. Under a multihomed environment, typically, a host has
more than one IP address at a time. During the transaction, the
host may be required to switch the IP address in use to another
IP address to preserve the communication. Such an address
update should be kept hidden from the upper layer protocols to
avoid communication disruption. The shim layer aims to make the
address update transparent to the upper layer protocols.</t>
<t>In a system which is based on identifier/locator separation,
upper layer protocols are expected to deal with identifiers for
establishing and handling the communications. If an application
wants to have multihoming support from the shim layer, the IP
addresses specified as source and destination addresses must be
identifiers. However, this does not necessarily mean that
applications are prohibited to choose specific locators for its
communication. It may be useful for some applications to
specify a preferred locator for a given flow.</t>
<t>This document recommends that the switching of identifier and
locator is done only once inside the TCP/IP stack of an endhost.
That is, if multiple shim sub-layers exist at the IP layer, any
one of them should be applied exclusively for a given flow.</t>
<t>As this document specifies sockets API extensions, it is
written so that the syntax and semantics are in line with the
Posix standard <xref target="POSIX"/> as much as possible. The
API specified in this document defines how to use ancillary data
(aka cmsg) to access the locator information with recvmsg()
and/or sendmsg() I/O calls. The definition of API is presented
in C language and data types follow the Posix format; intN_t
means a singed integer of exactly N bits (e.g. int16_t) and
uintN_t means an unsigned integer of exactly N bits
(e.g. uint32_t).</t>
<t>The target readers of this document are application
programmers who develop application software which may benefit
greatly from multihomed environments. In addition, this
document aims to provide necessary information for developers of
multihoming shim protocols to implement API for enabling
advanced locator management.</t>
</section>
<!--
================================================================
Terminology
================================================================
-->
<section title="Terminology" toc="include">
<t>This section provides terminology used in this document.
Basically most of the terms used in this document are taken from
the following documents:<vspace blankLines="1"/>
<list style="symbols">
<t>SHIM6 Protocol Specification<xref
target="I-D.ietf-shim6-proto"/></t>
<t>HIP Architecture<xref target="RFC4423"/></t>
<t>Reachability Protocol (REAP)<xref
target="I-D.ietf-shim6-failure-detection"/></t>
</list>
<vspace blankLines="1"/>
In this document, the term "IP" refers to both IPv4 and IPv6,
unless the protocol version is specifically mentioned. The
following are definitions of terms frequently used in this
document:<vspace blankLines="1"/>
<list style="symbols">
<t>Endpoint identifier (EID) - The identifier used by the
application to specify the endpoint of a given communication.
Applications may handle EIDs in various ways such as
long-lived connections, callbacks, and referrals<xref
target="I-D.ietf-shim6-app-refer"/>.
<list style="symbols">
<t>In the case of SHIM6, an identifier called a ULID serves
as an EID. A ULID is chosen from locators available on the
host.</t>
<t>In the case of HIP, an identifier called a Host
Identifier serves as an EID. A Host Identifier is derived
from the public key of a given host. For the sake of
backward compatibility with the sockets API, the Host
Identifier is represented in a form of hash of public key.
</t>
</list>
</t>
<t>Locator - The IP address actually used to deliver IP
packets. Locators are present in the source and destination
fields of the IP header of a packet on the wire. <list
style="symbols">
<t>List of locators - A list of locators associated with an
EID. There are two lists of locators stored in a given
context. One is associated with the local EID and the other
is associated with the remote EID. As defined in <xref
target="I-D.ietf-shim6-proto"/>, the list of locators
associated with an EID 'A' is denoted as Ls(A).</t>
<t>Preferred locator - The (source/destination) locator
currently used to send packets within a given context. As
defined in <xref target="I-D.ietf-shim6-proto"/>, the
preferred locator of a host 'A' is denoted as Lp(A).</t>
</list>
</t>
<t>Shim - The conceptual sub-layer inside the IP layer which
maintains mappings between EIDs and locators. An EID can be
associated with more than one locator at a time when the host
is multihomed. The term 'shim' does not refer to a specific
protocol but refers to the conceptual sub-layer inside the IP
layer.</t>
<t>Identifier/locator adaptation - The adaptation performed at
the shim layer which may end up re-writing the source and/or
destination addresses of an IP packet. In the outbound packet
processing, the EID pair is converted to the associated
locator pair. In the inbound packet processing, the locator
pair is converted to the EID pair.</t>
<t>Context - The state information shared by a given pair of
peers, which stores a binding between the EID and associated
locators. Contexts are maintained by the shim layer.</t>
<t>Reachability detection - The procedure to check
reachability between a given locator pair.</t>
<t>Path - The sequence of routers that an IP packet goes
through to reach the destination.</t>
<t>Path exploration - The procedure to explore available paths
for a given set of locator pairs.</t>
<t>Outage - The incident that prevents IP packets to flow from
the source locator to the destination locator. When there is
an outage, it means that there is no reachability between a
given locator pair. The outage may be caused by various
reasons, such as shortage of network resources, congestion,
and human error (faulty operation).</t>
<t>Working address pair - The address pair is considered to be
"working" if the packet can safely travel from the source to
the destination where the packet contains the first address
from the pair as the source address and the second address
from the pair as the destination address. If reachability is
confirmed in both directions, the address pair is considered
to be working bi-directionally.</t>
<t>Reachability protocol (REAP) - The protocol for detecting
failure and exploring reachability in a multihomed
environment. REAP is defined in <xref
target="I-D.ietf-shim6-failure-detection"/>.</t>
</list>
</t>
</section>
<!--
================================================================
System Overview
================================================================
-->
<section title="System Overview" anchor="sec-system-overview"
toc="include">
<t><xref target="fig-system-overview"/> illustrates the system
overview. The shim layer and REAP component exist inside the IP
layer. Applications use the sockets API defined in this
document to interface with the shim layer and the transport
layer for locator management, failure detection, and path
exploration.</t>
<t>It may also be possible that the shim layer interacts with
the transport layer, however, such an interaction is outside the
scope of this document.</t>
<figure anchor="fig-system-overview" title="System overview">
<artwork><![CDATA[
+------------------------+
| Application |
+------------------------+
^ ^
~~~~~~~~~~~~~|~Socket Interface|~~~~~~~~~~~~~~
| v
+-----------|------------------------------+
| | Transport Layer |
+-----------|------------------------------+
^ |
+-------------|-----|-------------------------------------+
| v v |
| +-----------------------------+ +----------+ | IP
| | Shim |<----->| REAP | | Layer
| +-----------------------------+ +----------+ |
| ^ ^ |
+-----------------------|----------------------|----------+
v v
+------------------------------------------+
| Link Layer |
+------------------------------------------+
]]></artwork>
</figure>
</section>
<!--
================================================================
Requirements
================================================================
-->
<section title="Requirements" anchor="sec-requirements" toc="include">
<t>The following is a list of requirements from applications:
<!--
<t>These requirements are mainly identified during the
discussions on SHIM6 WG mailing list. Some requirements
are derived from the REAP specification<xref
target="I-D.ietf-shim6-failure-detection"/>.<vspace
blankLines="1"/>
-->
<!--
Marcelo's email which was sent on the SHIM6 mailing list:
http://www.ops.ietf.org/lists/shim6/msg01191.html
-->
<list style="symbols">
<t>Locator management. The shim layer selects a pair of
locators for sending IP packets within a given context. The
selection is made by taking miscellaneous conditions into
account such as reachability of the path, application's
preference, and characteristics of path. From applications'
perspective:
<list style="symbols">
<t>It should be possible to obtain the lists of locators of
a given context: Ls(local) and Ls(remote).</t>
<t>It should be possible to obtain the preferred locators of
a given context: Lp(local) and Lp(remote).</t>
</list>
</t>
<t>Notification from applications to the shim layer about the
status of the communication. The notification occurs in an
event-based manner. Applications and/or upper layer protocols
may provide positive feedbacks or negative feedbacks to the
shim layer. [NOTE: These feedbacks are mentioned in <xref
target="I-D.ietf-shim6-failure-detection"/>]:
<list style="symbols">
<t>Applications and/or upper layer protocols (e.g., TCP) may
provide positive feedbacks to the shim layer informing that
the communication is going well.</t>
<t>Applications and/or upper layer protocols (e.g., TCP) may
provide negative feedbacks to the shim layer informing that
the communication status is not satisfactory. TCP may
detect a problem when it does not receive any expected ACK
message from the peer. Besides, a receipt of an ICMP error
message could be a clue for the application to detect
problems. The REAP module may be triggered by these
negative feedbacks and invoke the path exploration
procedure.</t> </list>
</t>
<t>Feedback from applications to the shim layer. Applications
should be able to inform the shim layer of the timeout values
for detecting failures, sending keepalives, and starting the
exploration procedure. In particular, applications should be
able to suppress keepalives.
</t>
<t>Hot-standby. Applications may request the shim layer for
the hot-standby capability. This means that, alternative
paths are known to be working in advance of a failure
detection. In such a case, it is possible for the host to
immediately replace the current locator pair with an
alternative locator pair.
</t>
<t>Eagerness for locator exploration. An application should
be able to inform the shim layer of how aggressively it wants
the REAP mechanism to perform a path exploration (e.g., by
specifying the number of concurrent attempts of discovery of
working locator pairs) when an outage occurs on the path
between the locator pair in use.</t>
<t>Providing locator information to applications. An
application should be able to obtain information about the
locator pair which was actually used to send or receive the
packet. <list style="symbols">
<t>For inbound traffic, the application may be interested in
the locator pair which was actually used to receive the
packet.
</t>
<t>For outbound traffic, the application may be interested
in the locator pair which was actually used to transmit the
packet.</t>
</list>
In this way, applications may have additional control on the
locator management. For example, an application becomes able
to verify if its preference for locator is actually applied to
the flow or not.
</t>
<t>Applications should be able to specify if they want to
defer the context setup, or if they want context establishment
to be started immediately in the case where there is no
available context. A deferred context setup means that the
initiation of communication should not be blocked to wait for
completion of the context establishment.</t>
<!-- turn on/off shim -->
<t>Turn on/off shim. An application should be able to request
to turn on or turn off the multihoming support by the shim
layer:
<list style="symbols">
<t>Apply shim. The application should be able to explicitly
request the shim layer to apply multihoming support.</t>
<t>Don't apply shim. The application should be able to
request the shim layer not to apply the multihoming support
but to apply normal IP processing at the IP layer.</t>
</list>
</t>
<t>An application should be able to know if the communication
is now being served by the shim layer or not.</t>
<t>An application should be able to use a common interface to
access an IPv4 locator and an IPv6 locator.</t>
</list>
</t>
</section>
<!--
===================================================================
Socket options for multihomed shim layer
===================================================================
-->
<section title="Socket Options for Multihoming Shim Layer"
anchor="sec-shim-socket-options" toc="include">
<t>In this section, socket options that are specific to multihomed
shim are defined.</t>
<t><xref target="tab-shim-socket-options"/> shows a list of the
socket options that are specific to the multihoming shim layer.
An application may specify these socket options for a given
socket either by the getsockopt() system call or by the
setsockopt() system call. All of these socket options are
defined at level SOL_SHIM.</t>
<t>The first column of <xref target="tab-shim-socket-options"/>
gives the name of the option. The second and third columns
indicate whether the option can be handled by the getsockopt()
system call and/or by the setsockopt() system call. The fourth
column provides a brief description of the socket option. The
fifth column shows the type of data structure specified along
with the socket option. By default, the data structure type is
an integer.</t>
<texttable anchor="tab-shim-socket-options"
title="Socket options for multihoming shim">
<ttcol align='left'>optname</ttcol>
<ttcol align='left'>get</ttcol>
<ttcol align='left'>set</ttcol>
<ttcol align='left'>description</ttcol>
<ttcol align='left'>dtype</ttcol>
<c>SHIM_ASSOCIATED</c>
<c>o</c>
<c></c>
<c>Check if the socket is associated with any shim context or
not.</c>
<c>int</c>
<c>SHIM_DONTSHIM</c>
<c>o</c>
<c>o</c>
<c>Request the shim layer not to apply any multihoming support
for the communication.</c>
<c>int</c>
<c>SHIM_HOT_STANDBY</c>
<c>o</c>
<c>o</c>
<c>Request the shim layer to prepare a hot-standby connection
(in addition to the current path).</c>
<c>int</c>
<c>SHIM_LOC_LOCAL_PREF</c>
<c>o</c>
<c>o</c>
<c>Get or set the preferred locator on the local side for the
context associated with the socket.</c>
<c>*1</c>
<c>SHIM_LOC_PEER_PREF</c>
<c>o</c>
<c>o</c>
<c>Get or set the preferred locator on the remote side for the
context associated with the socket.</c>
<c>*1</c>
<c>SHIM_LOC_LOCAL_RECV</c>
<c>o</c>
<c>o</c>
<c>Request for the destination locator of the received IP
packet.</c>
<c>int</c>
<c>SHIM_LOC_PEER_RECV</c>
<c>o</c>
<c>o</c>
<c>Request for the source locator of the received IP
packet.</c>
<c>int</c>
<c>SHIM_LOC_LOCAL_SEND</c>
<c>o</c>
<c>o</c>
<c>Request the use of specific locator as source locator of
outgoing IP packets.</c>
<c>*2</c>
<c>SHIM_LOC_PEER_SEND</c>
<c>o</c>
<c>o</c>
<c>Request the use of specific locator as destination locator
of outgoing IP packets.</c>
<c>*2</c>
<c>SHIM_LOCLIST_LOCAL</c>
<c>o</c>
<c>o</c>
<c>Get or set the list of locators associated with the local
EID.</c>
<c>*3</c>
<c>SHIM_LOCLIST_PEER</c>
<c>o</c>
<c>o</c>
<c>Get or set the list of locators associated with the peer's
EID.</c>
<c>*3</c>
<c>SHIM_APP_TIMEOUT</c>
<c>o</c>
<c>o</c>
<c>Inform the shim layer of the timeout value for detecting
failure.</c>
<c>int</c>
<c>SHIM_PATHEXPLORE</c>
<c>o</c>
<c>o</c>
<!-- marcelo: not sure what this means... is the number of
alternative paths to be explored, how many times each
path will be tried i.e. how many probe packets per
alternative path are to be send? -->
<c>Specify behavior of path exploration and failure
detection.</c>
<c>*4</c>
<c>SHIM_CONTEXT_DEFERRED_SETUP</c>
<c>o</c>
<c>o</c>
<c>Specify if the context setup can be deferred or not.</c>
<c>int</c>
</texttable>
<t>*1: Pointer to a shim_locator which is defined in <xref
target="sec-data-structures"/>.</t>
<t>*2: Pointer to shim_locator data structure.</t>
<t>*3: Pointer to an array of shim_locator.</t>
<t>*4: Pointer to a shim_pathexplore which is defined in <xref
target="sec-data-structures"/>.</t>
<t><xref target="fig-socket-api-model"/> illustrates how the
shim specific socket options fit into the system model of socket
API. The figure shows that the shim layer and the additional
protocol components (IPv4 and IPv6) below the shim layer are new
to the system model. As previously mentioned, all the shim
specific socket options are defined at SOL_SHIM level. This
design choice brings the following advantages:<vspace
blankLines="1"/>
<list style="numbers">
<t>The existing sockets API continue to work at the layer
above the shim layer. That is, those legacy API handle IP
addresses as identifiers.</t>
<t>With newly defined socket options for the shim layer, the
application obtains additional control of locator
management.</t>
<t>The shim specific socket options can be kept independent
from address family (IPPROTO_IP or IPPROTO_IPV6) and transport
protocol (IPPROTO_TCP or IPPROTO_UDP).</t> </list>
</t>
<figure anchor="fig-socket-api-model"
title="System model of sockets API with shim layer">
<artwork><![CDATA[
s1 s2 s3 s4
| | | |
+----------------|--|-------|--|----------------+
| +-------+ +-------+ |
| IPPROTO_TCP | TCP | | UDP | |
| +-------+ +-------+ |
| | \ / | |
| | ----- | |
| | / \ | |
| +------+ +------+ |
| IPPROTO_IP | IPv4 | | IPv6 | IPPROTO_IPV6 |
| +------+ +------+ |
| \ / SOL_SOCKET
| +--------\-------/--------+ |
| SOL_SHIM | shim | |
| +--------/-------\--------+ |
| / \ |
| +------+ +------+ |
| | IPv4 | | IPv6 | |
| +------+ +------+ |
| | | |
+------------------|----------|-----------------+
| |
IPv4 IPv6
Datagram Datagram
]]></artwork>
</figure>
<section title="SHIM_ASSOCIATED">
<t>The SHIM_ASSOCIATED option can be used to check whether the
socket is associated with any shim context or not.</t>
<t>This option is particularly meaningful in the case where
the locator information of the received IP packet does not
tell whether the identifier/locator adaptation is performed or
not. Note that the EID pair and locator pair may be identical
in some case.</t>
<t>This option can be specified by getsockopt(). Thus, the
option is read-only and the result (0 or 1) is set in the
option value (the fourth argument of getsockopt()).</t>
<t>The data type of the option value is an integer. The
option value indicates the presence of shim context. A
returned value 1 means that the socket is associated with a
shim context at the shim layer, while a return value 0
indicates that there is no shim context associated with the
socket.</t>
<t>For example, the option can be used by the application as
follows:</t>
<figure>
<artwork><![CDATA[
int optval;
int optlen = sizeof(optval);
getsockopt(fd, SOL_SHIM, SHIM_ASSOCIATED, &optval, &optlen);
]]></artwork>
</figure>
</section>
<section title="SHIM_DONTSHIM">
<t>The SHIM_DONTSHIM option can be used to request the shim
layer to not apply the multihoming support for the
communication established over the socket.</t>
<t>The data type of the option value is an integer. The
option value indicates whether the multihoming shim support is
deprecated or not. The option value is binary (0 or 1). By
default, the value is set to 0, which means that the shim
layer applies identifier/locator adaptation for the flow. In
order to disable the socket option, the application should
call setsockopt() with optval set to 0.</t>
<t>For example, the application may disable the socket option
as follows:</t>
<figure>
<artwork><![CDATA[
int optval;
optval = 0;
setsockopt(fd, SOL_SHIM, SHIM_DONTSHIM, &optval, sizeof(optval));
]]></artwork>
</figure>
<t>For example, the application may check the option value as
follows:</t>
<figure>
<artwork><![CDATA[
int optval;
int len;
len = sizeof(optval);
getsockopt(fd, SOL_SHIM, SHIM_DONTSHIM, &optval, &len);
]]></artwork>
</figure>
</section>
<section title="SHIM_HOT_STANDBY">
<t>The SHIM_HOT_STANDBY option can be used to check if the
shim layer uses hot-standby connection or not for the
communication established over the socket. A hot-standby
connection is based on an alternative working locator pair to
the current locator pair. This option is effective only when
there is a shim context associated with the socket.</t>
<t>The data type of the option value is an integer.</t>
<t>The option value can be set by setsockopt().</t>
<t>The option value can be read by getsockopt().</t>
<t>By default, the value is set to 0, meaning that hot-standby
connection is disabled.</t>
<t>For example, the option can be activated by the application
as follows.</t>
<figure>
<artwork><![CDATA[
int optval;
optval = 1;
setsockopt(fd, SOL_SHIM, SHIM_HOT_STANDBY, &optval,
sizeof(optval));
]]></artwork>
</figure>
<t>For example, the option value can be checked by the
application as follows.</t>
<figure>
<artwork><![CDATA[
int optval;
int len;
len = sizeof(optval);
getsockopt(fd, SOL_SHIM, SHIM_HOT_STANDBY, &optval, &len);
]]></artwork>
</figure>
</section>
<section title="SHIM_PATHEXPLORE">
<t> The application may specify this socket option to specify
specify behavior of path exploration. Path exploration is a
procedure to find an alternative locator pair when the host
finds any problem with the current locator pair. The message
used for finding an alternative locator pair is called the
Probe message and it is sent per locator pair. The REAP
specification defines the default values for Initial Probe
Timeout and Initial Probe.</t>
<!--
Initial Probe Timeout = 0.5 seconds
Initial Probe = 4 times
-->
<t>The option is effective only when there is a shim context
associated with the socket.</t>
<t>The data type of the option value is a pointer to the
buffer where a set of information for path exploration is
stored. The data structure is defined in <xref
target="sec-data-structures"/>.</t>
<t>By default, the option value is set to NULL, meaning that
the option is disabled.</t>
<t>An error ENOENT will be returned when there is no context
associated with the socket.</t>
<t>For example, the parameters for the path exploration can be
set as follows.</t>
<figure>
<artwork><![CDATA[
struct shim6_pathexplore pe;
pe.pe_probenum = 4; /* times */
pe.pe_keepaliveto = 10; /* seconds */
pe.pe_initprobeto = 500; /* milliseconds */
pe.pe_reserved = 0;
setsockopt(fd, SOL_SHIM, SHIM_PATHEXPLORE, &pe, sizeof(pe));
]]></artwork>
</figure>
<t>For example, the parameters for the path exploration can be
read as follows.</t>
<figure>
<artwork><![CDATA[
struct shim6_pathexplore pe;
int len;
len = sizeof(pe);
getsockopt(fd, SOL_SHIM, SHIM_PATHEXPLORE, &pe, &len);
]]></artwork>
</figure>
</section>
<section title="SHIM_LOC_LOCAL_PREF">
<t>The SHIM_LOC_LOCAL_PREF option can be used to read or set
preferred locator on local side within a given context. Hence
this option is effective only when there is a shim context
associated with the socket.</t>
<t>The data type of the option value is a pointer to the
specific data structure which stores the locator information.
The data structure is defined in <xref
target="sec-data-structures"/>.</t>
<t>By default, the option value is set to NULL, meaning that
the option is disabled.</t>
<t>The preferred locator can be set by setsockopt().
Verification of the locator shall be done by the shim layer
before updating the preferred locator.</t>
<t>The preferred locator can be read by getsockopt().</t>
<t>An error ENOENT will be returned when there is no context
associated with the socket.</t>
<t>An error EINVALIDLOCATOR will be returned when the
validation of the specified locator failed.</t>
<t>For example, a preferred locator can be set as follows. It
should be noted that some members of the shim_locator
(lc_ifidx and lc_flags) are ignored in the write
operation.</t>
<figure>
<artwork><![CDATA[
struct shim_locator lc;
struct in6_addr ip6;
/* ...set the locator (ip6)... */
memset(&lc, 0, sizeof(shim_locator));
lc.lc_family = AF_INET6; /* IPv6 */
lc.lc_ifidx = 0;
lc.lc_flags = 0;
lc.lc_preference = 255;
memcpy(lc.lc_addr, &ip6, sizeof(in6_addr));
setsockopt(fd, SOL_SHIM, SHIM_LOC_LOCAL_PREF, &lc,
sizeof(optval));
]]></artwork>
</figure>
<t>For example, the preferred locator of the context can be
read by application as follows.</t>
<figure>
<artwork><![CDATA[
struct shim_locator lc;
int len;
len = sizeof(lc);
getsockopt(fd, SOL_SHIM, SHIM_LOC_LOCAL_PREF, &lc, &len);
]]></artwork>
</figure>
</section>
<section title="SHIM_LOC_PEER_PREF">
<t>The SHIM_LOC_PEER_PREF option can be used to read or set
preferred locator on peer side within a given context. Hence
this option is effective only when there is a shim context
associated with the socket.</t>
<t>The data type of the option value is a pointer to the
specific data structure which stores the locator information.
The data structure is defined in <xref
target="sec-data-structures"/>.</t>
<t>By default, the option value is set to NULL, meaning that
the option is disabled.</t>
<t>The preferred locator can be set by setsockopt(). The shim
layer shall perform verification of the locator before
updating the preferred locator.</t>
<t>The preferred locator can be read by getsockopt().</t>
<t>An error ENOENT will be returned when there is no context
associated with the socket.</t>
<t>An error EINVALIDLOCATOR will be returned when the
validation of the specified locator failed.</t>
<t>For example, a preferred locator can be set as follows. It
should be noted that some members of the shim_locator
(lc_ifidx and lc_flags) are ignored in the write
operation.</t>
<t>The usage of the option is same as that of
SHIM_LOC_LOCAL_PREF.</t>
</section>
<section title="SHIM_LOC_LOCAL_RECV">
<t>The SHIM_LOC_LOCAL_RECV option can be used to request the
shim layer to store the destination locator of the received IP
packet in an ancillary data object which can be accessed by
recvmsg(). Hence this option is effective only when there is
a shim context associated with the socket.</t>
<t>The data type of the option value is integer. The option
value should be binary (0 or 1). By default, the option value
is set to 0, meaning that the option is disabled.</t>
<t>The option value can be set by setsockopt().</t>
<t>The option value can be read by getsockopt().</t>
<t>See <xref target="sec-access-to-locinfo"/> for the
procedure to access locator information stored in the
ancillary data objects.</t>
<t>An error ENOENT will be returned when there is no context
associated with the socket.</t>
<t>For example, the option can be activated by the application
as follows:</t>
<figure>
<artwork><![CDATA[
int optval;
optval = 1;
setsockopt(fd, SOL_SHIM, SHIM_LOC_LOCAL_RECV, &optval,
sizeof(optval));
]]></artwork>
</figure>
<t>For example, the option value can be checked by the
application as follows:</t>
<figure>
<artwork><![CDATA[
int optval;
int len;
len = sizeof(optval);
getsockopt(fd, SOL_SHIM, SHIM_LOC_LOCAL_RECV, &optval, &len);
]]></artwork>
</figure>
</section>
<section title="SHIM_LOC_PEER_RECV">
<t>The SHIM_LOC_PEER_RECV option can be used to request the
shim layer to store the source locator of the received IP
packet in an ancillary data object which can be accessed by
recvmsg(). Hence this option is effective only when there is
a shim context associated with the socket.</t>
<t>The data type of the option value is integer. The option
value should be binary (0 or 1). By default, the option value
is set to 0, meaning that the option is disabled.</t>
<t>The option value can be set by setsockopt().</t>
<t>The option value can be read by getsockopt().</t>
<t>See <xref target="sec-access-to-locinfo"/> for the
procedure to access locator information stored in the
ancillary data objects.</t>
<t>An error ENOENT will be returned when there is no context
associated with the socket.</t>
<t>The usage of the option is same as that of
SHIM_LOC_LOCAL_RECV option.</t>
</section>
<section title="SHIM_LOC_LOCAL_SEND">
<t>The SHIM_LOC_LOCAL_SEND option can be used to request the
shim layer to use specific locator for the source locator of
IP packets to be sent from the socket. Hence this option is
effective only when there is a shim context associated with
the socket.</t>
<t>The data type of option value is pointer to shim_locator
data structure.</t>
<t>The local locator can be specified by setsockopt()
providing a valid locator which is stored in a shim_locator
data structure. When a zero-filled locator is specified,
pre-existing setting of local locator is inactivated.</t>
<t>The local locator specified can be obtained by
getsockopt(). The locator can be obtained from the option
value.</t>
<t>An error ENOENT will be returned when there is no context
associated with the socket.</t>
<t>An error EINVALIDLOCATOR when invalid locator is
specified.</t>
<t>For example, a preferred local locator can be specified as
follows.</t>
<figure>
<artwork><![CDATA[
struct shim_locator locator;
struct in6_addr ia6;
/* an IPv6 address preferred for the source locator is copied
to the parameter ia6 */
memset(&locator, 0, sizeof(locator));
/* fill shim_locator data structure */
locator.lc_family = AF_INET6;
locator.lc_ifidx = 1;
locator.lc_flags = 0;
locator.lc_preference = 0;
memcpy(&locator.lc_addr, &ia6, sizeof(ia6));
setsockopt(fd, SOL_SHIM, SHIM_LOC_LOCAL_SEND, &locator,
sizeof(locator));
]]></artwork>
</figure>
<t>For example, a preferred local locator can be read as
follows.</t>
<figure>
<artwork><![CDATA[
struct shim_locator locator;
memset(&locator, 0, sizeof(locator));
getsockopt(fd, SOL_SHIM, SHIM_LOC_LOCAL_SEND, &locator,
sizeof(locator));
/* check locator */
]]></artwork>
</figure>
</section>
<section title="SHIM_LOC_PEER_SEND">
<t>The SHIM_LOC_PEER_SEND option can be used to request the
shim layer to use specific locator for the destination locator
of IP packets to be sent from the socket. Hence this option
is effective only when there is a shim context associated with
the socket.</t>
<t>The data type of the option value is a pointer to
shim_locator data structure.</t>
<t>The remote locator can be specified by setsockopt()
providing a valid locator which is stored in a shim_locator
data structure. When a zero-filled locator is specified,
pre-existing setting of remote locator is inactivated.</t>
<t>The remote locator specified can be obtained by
getsockopt(). The locator can be obtained from the option
value.</t>
<t>An error ENOENT will be returned when there is no context
associated with the socket.</t>
<t>An error EINVALIDLOCATOR when invalid locator is
specified.</t>
<t>The usage of the option is as the same as that of
SHIM_LOC_LOCAL_SEND option.</t>
</section>
<section title="SHIM_LOCLIST_LOCAL">
<t>The SHIM_LOCLIST_LOCAL option can be used to read or set
the locator list associated with the local EID of the shim
context associated with the socket. Hence this option is
effective only when there is a shim context associated with
the socket.</t>
<t>The data type of the option value is a pointer to the
buffer where a locator list is stored. See <xref
target="sec-data-structures"/> for the data structure for
storing the locator information. By default, the option value
is set to NULL, meaning that the option is disabled.</t>
<t>The locator list can be read by getsockopt(). Note that
the size of the buffer pointed by optval argument should be
large enough to store an array of locator information. The
number of the locator information is not known beforehand.</t>
<t>The locator list can be set by setsockopt(). The buffer
pointed by optval argument should contain an array of locator
list.</t>
<t>An error ENOENT will be returned when there is no context
associated with the socket.</t>
<t>An error EINVALIDLOCATOR will be returned when the
validation of the specified locator failed.</t>
<!-- xxx do we need more error code? -->
<t>For example, a list of locators to be associated with the
local EID can be specified as follows:</t>
<figure>
<artwork><![CDATA[
struct shim_locator locators[SHIM_MAX_LOCATORS];
struct sockaddr_in *sin;
struct sockaddr_in6 *sin6;
memset(locators, 0, sizeof(locators));
...
/* obtain local IP addresses from local interfaces */
...
/* first locator (an IPv6 address) */
locators[0].lc_family = AF_INET6;
locators[0].lc_ifidx = 0;
locators[0].lc_flags = 0;
locators[0].lc_preference = 1;
memcpy(&locators[0].lc_addr, &sa6->sin6_addr,
sizeof(sa6->sin6_addr));
...
/* second locator (an IPv4 address) */
locators[1].lc_family = AF_INET;
locators[1].lc_ifidx = 0;
locators[1].lc_flags = 0;
locators[1].lc_preference = 0;
memcpy(&locators[1].lc_addr, &sa->sin_addr, sizeof(sa->sin_addr));
setsockopt(fd, SOL_SHIM, SHIM_LOCLIST_LOCAL, locators,
sizeof(locators));
]]></artwork>
</figure>
<t>For example, a list of locators that are associated with the
local EID can be obtained as follows:</t>
<figure>
<artwork><![CDATA[
struct shim_locator locators[SHIM_MAX_LOCATORS];
memset(locators, 0, sizeof(locators));
getsockopt(fd, SOL_SHIM, SHIM_LOC_LOCAL_RECV, locators,
sizeof(locators));
/* parse locators */
...
]]></artwork>
</figure>
</section>
<section title="SHIM_LOCLIST_PEER">
<t>The SHIM_LOCLIST_PEER option can be used to read or set the
locator list associated with the peer EID of the shim context
associated with the socket. Hence this option is effective
only when there is a shim context associated with the
socket.</t>
<t>The data type of the option value is a pointer to the
buffer where a locator list is stored. See <xref
target="sec-data-structures"/> for the data structure for
storing the locator information. By default, the option value
is set to NULL, meaning that the option is disabled.</t>
<t>The locator list can be read by getsockopt(). Note that
the size of the buffer pointed by optval argument should be
large enough to store an array of locator information. The
number of the locator information is not known beforehand.</t>
<t>The locator list can be set by setsockopt(). The buffer
pointed by optval argument should contain an array of locator
list.</t>
<t>An error ENOENT will be returned when there is no context
associated with the socket.</t>
<t>An error EINVALIDLOCATOR will be returned when the
validation of the specified locator failed.</t>
<!-- xxx do we need more error code? -->
<t>The usage of the option is same as that of
SHIM_LOCLIST_LOCAL.</t>
</section>
<section title="SHIM_APP_TIMEOUT">
<t>The SHIM_APP_TIMEOUT option indicates timeout value for
application to detect failure. Hence this option is effective
only when there is a shim context associated with the
socket.</t>
<t>The data type of the option value is an integer. The value
indicates the period of timeout in seconds to send a REAP
Keepalive message since the last outbound traffic. By
default, the option value is set to 0, meaning that the option
is disabled. When the option is disabled, the REAP mechanism
follows its default value of Send Timeout value as specified
in <xref target="I-D.ietf-shim6-failure-detection"/></t>
<t>If the timeout value specified is longer than the Send
Timeout configured in the REAP component, the REAP Keepalive
message should be suppressed.</t>
<t>An error ENOENT will be returned when there is no context
associated with the socket.</t>
<t>For example, a specific timeout value can be configured by
the application as follows:</t>
<figure>
<artwork><![CDATA[
int optval;
optval = 15; /* 15 seconds */
setsockopt(fd, SOL_SHIM, SHIM_APP_TIMEOUT, &optval,
sizeof(optval));
]]></artwork>
</figure>
<t>For example, the option value namely the period of timeout
can be checked by the application as follows:</t>
<figure>
<artwork><![CDATA[
int optval;
int len;
len = sizeof(optval);
getsockopt(fd, SOL_SHIM, SHIM_APP_TIMEOUT, &optval, &len);
]]></artwork>
</figure>
</section>
<section title="SHIM_DEFERRED_CONTEXT_SETUP">
<t>The SHIM_DEFERRED_CONTEXT_SETUP option indicates how
initiation of context setup is made in terms of timing (before
or after) the initial communication flow. Deferred context
means that the establishment of context does not put
additional delay for an initial transaction.</t>
<t>The data type for the option value is an integer. The
option value should binary (0 or 1). By default, the value is
set to 1, meaning that the context setup is deferred. In
order to disable the option, the application should call
setsockopt() with option value set to 0.</t>
<t>However, it should be noted that deferred context setup may
not be possible in some cases. For instance, an EID may be
non-routable address (e.g., Host Identifier in HIP) and there
is no way to transmit any IP packet unless there is a context
providing the locators. In such a case, a context should be
established prior to the communication.</t>
<t>For example, the option can be disabled by the application
as follows:</t>
<figure>
<artwork><![CDATA[
int optval;
optval = 0;
setsockopt(fd, SOL_SHIM, SHIM_DEFERRED_CONTEXT_SETUP,
&optval, sizeof(optval));
]]></artwork>
</figure>
<t>For example, the option value can be checked by the
application as follows:</t>
<figure>
<artwork><![CDATA[
int optval;
int len;
len = sizeof(optval);
getsockopt(fd, SOL_SHIM, SHIM_DEFERRED_CONTEXT_SETUP,
&optval, &len);
]]></artwork>
</figure>
</section>
<section title="Error Handling">
<t>If successful, getsockopt() and setsockopt() return 0;
otherwise, the functions return -1 and set errno to indicate
error.</t>
<t>The following are new error values defined for some shim
specific socket options indicating that the getsockopt() or
setsockopt() finished incompletely:<vspace blankLines="1"/>
<list style="hanging">
<t hangText="EINVALIDLOCATOR"><vspace blankLines="0"/>This
indicates that at least one of the necessary validations
inside the shim layer for the specified locator has failed.
In case of SHIM6, there are two kinds of verifications
required for security reasons prior to sending an IP packet
to the peer's new locator; one is the return routability
(check if the peer is actually willing to receive data with
the specified locator) and the other one is the verification
based on crypto identifier mechanisms <xref
target="RFC3972"/>, <xref target="I-D.ietf-shim6-hba"/>.</t>
</list>
</t>
</section>
</section>
<section title="Ancillary Data for Multihoming Shim"
anchor="sec-access-to-locinfo" toc="include">
<t>In this section, the definition and the usage of the
ancillary data specific to multihoming shim are provided.</t>
<t>As defined in the Posix standard, sendmsg() and recvmsg()
input a msghdr structure as their arguments. These system calls
can handle control information along with data. <xref
target="fig-msghdr"/> shows the msghdr structure which is
defined in <sys/socket.h>. The member msg_control holds a
pointer to the buffer where the shim specific ancillary data
objects can be stored in addition to other ancillary data
objects.</t>
<figure anchor="fig-msghdr" title="msghdr structure">
<artwork><![CDATA[
struct msghdr {
caddr_t msg_name; /* optional address */
u_int msg_namelen; /* size of address */
struct iovec *msg_iov; /* scatter/gather array */
u_int msg_iovlen; /* # elements in msg_iov */
caddr_t msg_control; /* ancillary data, see below */
u_int msg_controllen; /* ancillary data buffer len */
int msg_flags; /* flags on received message */
};
]]></artwork>
</figure>
<t>The buffer pointed by the member msg_control of the msghdr
structure may contain locator information which is a single
locator and it should be possible to process them with the
existing macros defined in Posix and <xref target='RFC3542'/>.
Each cmsghdr{} should be followed by data which stores a single
locator.</t>
<t>In case of non-connected socket, msg_name member stores the
socket address of the peer which should be considered as an
identifier rather than a locator. The locator of the peer node
should be retrieved by SHIM_LOC_PEER_RECV as specified
below.</t>
<t><xref target="tab-shim-ancillary-data"/> is a list of the
shim specific ancillary data which can be used for recvmsg() or
sendmsg(). In any case, SOL_SHIM must be set as cmsg_level.</t>
<texttable anchor="tab-shim-ancillary-data"
title="Shim specific ancillary data">
<ttcol align='left'>cmsg_type</ttcol>
<ttcol align='left'>sendmsg()</ttcol>
<ttcol align='left'>recvmsg()</ttcol>
<ttcol align='left'>cmsg_data[]</ttcol>
<c>SHIM_LOC_LOCAL_RECV</c>
<c></c>
<c>o</c>
<c>*1</c>
<c>SHIM_LOC_PEER_RECV</c>
<c></c>
<c>o</c>
<c>*1</c>
<c>SHIM_LOC_LOCAL_SEND</c>
<c>o</c>
<c></c>
<c>*1</c>
<c>SHIM_LOC_PEER_SEND</c>
<c>o</c>
<c></c>
<c>*1</c>
<c>SHIM_FEEDBACK</c>
<c>o</c>
<c></c>
<c>shim_feedback{}</c>
</texttable>
<t>*1: cmsg_data[] should include padding (if necessary) and a
single sockaddr_in{}/sockaddr_in6{}.</t>
<t>It should be noted that the above ancillary data can only be
handled by a UDP or a raw socket and not by a TCP socket. This
is because there is no one-to-one mapping between a single
send/receive operation and a TCP segment being
transmitted/received.</t>
<section title="Get Locator Information from Incoming Packet">
<t>An application can get locator information from the
received IP packet by specifying the shim specific socket
options for the socket. When SHIM_LOC_LOCAL_RECV and/or
SHIM_LOC_PEER_RECV socket options are set, the application can
retrieve local and/or remote locator from the ancillary
data.</t>
</section>
<section title="Specify Locator Information for Outgoing Packet">
<t>An application can specify the locators to be used for
transmitting an IP packet by sendmsg(). When the ancillary
data of cmsg_type SHIM_LOC_LOCAL_SEND and/or
SHIM_LOC_PEER_SEND are specified, the application can
explicitly specify the source and/or the destination locators
to be used for the communication over the socket.</t>
<t>In addition, the application can specify the outgoing
interface by SHIM_IF_SEND ancillary data. The ancillary data
should contain the interface identifier of the physical
interface over which the application expects the packet to be
transmitted.</t>
<t>Note that the effect is limited to the datagram transmitted
by the sendmsg().</t>
<t>If the specified locator pair is verified, the shim layer
overrides the locators of the IP packet.</t>
<t>An error EINVALIDLOCATOR will be returned when validation
of the specified locator failed.</t>
</section>
<section title="Notification from Application to Multihoming Shim"
anchor="sec-feedback" toc="include">
<t>An application may provide feedbacks to the shim layer
about the communication status. Such feedbacks are
particularly useful for the shim layer in the absence of REAP
mechanism to monitor the reachability status of the currently
used locator pair in a given shim context.</t>
<t>The notification can be made by sendmsg() specifying a new
ancillary data called SHIM_FEEDBACK. The ancillary data can
be handled by specifying SHIM_FEEDBACK option in
cmsg_type.</t>
<t>An error ENOENT will be returned when there is no context
associated with the socket.</t>
<t>See <xref target="sec-feedback-info"/> for details of the
data structure to be used. Note that this specification does
not specify the exact behavior of the shim layer when a
feedback is given by an application.</t>
</section>
</section>
<section title="Data Structures"
anchor="sec-data-structures" toc="include">
<t>In this section, data structures specifically defined for the
multihoming shim layer are introduced. These data structures
are either used as a parameter for setsockopt()/getsockopt() (as
mentioned in <xref target="sec-shim-socket-options"/>) or as a
parameter for ancillary data to be processed by
sendmsg()/recvmsg() (as mentioned in <xref
target="sec-access-to-locinfo"/>).</t>
<!--
Note: maybe we can define getlocatorinfo() system call which
returns chain of locator information associated with a given
identifier. For instance: getlocatorinfo(const struct
sockaddr *sa_id, const struct addrinfo *hints, struct
addrinfo **result);
-->
<section title="Placeholder for Locator Information">
<t>As defined in <xref target="sec-shim-socket-options"/>, the
SHIM_LOC_LOCAL_PREF, SHIM_LOC_PEER_PREF, SHIM_LOCLIST_LOCAL,
and SHIM_LOCLIST_PEER socket options need to handle one or
more locator information. Locator information includes not
only the locator itself but also additional information about
the locator which is useful for locator management. A new
data structure is defined to serve as a placeholder for the
locator information.</t>
<t><xref target="fig-shim-locator"/> illustrates the data
structure called shim_locator which stores a locator
information.
<figure anchor="fig-shim-locator" title="shim locator structure">
<artwork><![CDATA[
struct shim_locator {
uint8_t lc_family; /* address family */
uint8_t lc_ifidx; /* interface index */
uint8_t lc_flags; /* flags */
uint8_t lc_preference; /* preference value */
uint8_t lc_addr[16]; /* address data */
};
]]></artwork>
</figure>
<list style="hanging">
<t hangText="lc_family"><vspace blankLines="0"/>Address
family of the locator (e.g. AF_INET, AF_INET6). It is
required that the parameter contains non-zero value
indicating the exact address family of the locator.</t>
<t hangText="lc_ifidx"><vspace blankLines="0"/>Interface
index of the network interface to which the locator is
assigned. This field should be valid only in a read
(getsockopt()) operation.</t>
<t hangText="lc_flags"><vspace blankLines="0"/>Each bit of
the flags represents a specific characteristics of the
locator. Hash Based Address (HBA) is defined as 0x01.
Cryptographically Generated Address (CGA) is defined as
0x02.
</t>
<t hangText="lc_preference"><vspace
blankLines="0"/>Indicates a preference of the locator. The
preference is represented by an integer.</t> <t
hangText="lc_addr"><vspace blankLines="0"/>Contains the
locator. In the case where a locator whose size is smaller
than 16 bytes, an encoding rule should be provided for each
locator of a given address family. For instance, in case of
AF_INET (IPv4), the locator should be in the format of an
IPv4-mapped IPv6 address as defined in RFC 4291<xref
target="RFC4291"/>.</t></list>
</t>
</section>
<section title="Path Exploration Parameter">
<t>As defined in <xref target="sec-shim-socket-options"/>,
SHIM_PATHEXPLORE allows application to set or read the
parameters for path exploration and failure detection. A new
data structure called shim_pathexplore is defined to store the
necessary parameters. <xref target="fig-path-explore"/>
illustrates the data structure. The data structure can be
passed to getsockopt() or setsockopt() as an argument.
<figure anchor="fig-path-explore" title="path explore structure">
<artwork><![CDATA[
struct shim_pathexplore {
uint8_t pe_probenum; /* # of initial probe */
uint8_t pe_keepaliveto; /* Keepalive Timeout */
uint16_t pe_initprobeto; /* Initial Probe Timeout */
uint32_t pe_reserved; /* reserved */
};
]]></artwork>
</figure>
<list style="hanging">
<t hangText="pe_probenum">
<vspace blankLines="0"/>
Indicates the number of initial probe messages to be sent.
Default value of this parameter should follow what is
specified in <xref
target="I-D.ietf-shim6-failure-detection"/>.
</t>
<t hangText="pe_keepaliveto">
<vspace blankLines="0"/>
Indicates timeout value for detecting a failure when the
host does not receive any packets for a certain period of
time while there is outbound traffic. When the timer
expires, path exploration procedure will be carried out by
sending a REAP Probe message. Default value of this
parameter should follow what is specified in <xref
target="I-D.ietf-shim6-failure-detection"/>.
</t>
<t hangText="pe_initprobeto">
<vspace blankLines="0"/>
Indicates retransmission timer of REAP Probe message in
milliseconds. Note that this timer is applied before
exponential back-off is started. A REAP Probe message for
the same locator pair may be retransmitted. Default value
of this parameter should follow what is specified in <xref
target="I-D.ietf-shim6-failure-detection"/>.
</t>
<t hangText="pe_reserved">
<vspace blankLines="0"/>
A reserved field for future extension. By default, the
field should be initialized to zero.
</t>
</list>
</t>
</section>
<section title="Feedback Information" anchor="sec-feedback-info">
<t>As mentioned in <xref target="sec-feedback"/>, applications
can inform the shim layer about the status of unicast
reachability of the locator pair currently in use. The
feedback information can be handled by using ancillary data
called SHIM_FEEDBACK. A new data structure named
shim_feedback is illustrated in <xref
target="fig-feedback-info"/>.
<figure anchor="fig-feedback-info"
title="feedback information structure">
<artwork><![CDATA[
struct shim_feedback {
uint8_t fb_direction; /* direction of traffic */
uint8_t fb_indicator; /* indicator (1-3) */
uint16_t fb_reserved; /* reserved */
};
]]></artwork>
</figure>
<list style="hanging">
<t hangText="direction">
<vspace blankLines="0"/>
Indicates direction of reachability between a locator pair
in question. A value 0 indicates outbound and a value 1
indicates inbound direction.
</t>
<t hangText="indicator">
<vspace blankLines="0"/>
A value indicating the degree of satisfaction of a
unidirectional reachability for a given locator pair.
<list style="symbols">
<t>0: Default value. Whenever this value is specified
the feedback information must not be processed by the
shim layer.</t>
<t>1: Unable to connect. There is no unidirectional
reachability between the locator pair in question.</t>
<t>2: Unsatisfactory. The application is not satisfied
with the unidirectional reachability between the locator
pair in question.</t>
<t>3: Satisfactory. There is satisfactory
unidirectional reachability between the locator pair in
question.</t>
</list>
</t>
<t hangText="reserved">
<vspace blankLines="0"/>
Reserved field. Must be ignored by the receiver.
</t>
</list>
</t>
</section>
</section>
<section title="Implications for Existing Socket API Extensions"
anchor="sec-implications-for-legacyapi"
toc="include">
<t>Some of the socket options defined in this document are
overlapping with existing sockets API and care should be taken
for the usage not to confuse with the overlapping features.</t>
<t>The socket options for requesting specific locators to be
used for a given transaction (SHIM_LOC_LOCAL_PREF and
SHIM_LOC_PEER_PREF) are semantically similar to the existing
sockets API (IPV6_PKTINFO). The socket options for obtaining
the locator information from the received IP packet
(SHIM_LOC_LOCAL_RECV and SHIM_LOC_PEER_RECV) are semantically
similar to the existing sockets API (IP_RECVDSTADDR and
IPV6_PKTINFO).</t>
<t>In IPv4, application can obtain the destination IP address of
the received IP packet (IP_RECVDSTADDR). If the shim layer
performs identifier/locator adaptation for the received packet,
the destination EID should be stored in the ancillary data
(IP_RECVDSTADDR).</t>
<t>In IPv6, <xref target="RFC3542"/> defines that IPV6_PKTINFO
can be used to specify source IPv6 address and the outgoing
interface for outgoing packets, and retrieve destination IPv6
address and receiving interface for incoming packets. This
information is stored in ancillary data being IPV6_PKTINFO
specified as cmsg_type. Existing sockets API should continue to
work above the shim layer, that is, the IP addresses handled in
IPV6_PKTINFO should be EIDs, not the locators.</t>
<t>Baseline is that the above existing sockets API
(IP_RECVDSTADDR and IPV6_PKTINFO) is assumed to work above the
multihoming shim layer. In other words, the IP addresses those
socket options deal with are EIDs rather than locators.</t>
</section>
<section title="Resolving Conflicts with Preference Values">
<t>Since the multihoming shim API allows application to specify
preference value for the context which is associated with the
socket instance, there may be a conflict with preference values
specified by different applications. For instance, application
A and B may establish communication over the same EID pair while
both applications have different preference in their choice of
local locator.</t>
<t>SHIM6 supports a notion of context forking in which a context
is split when there is a conflict with preference values
specified by multiple applications. Thus, context forking can
simply resolve the conflicting situation which may be caused by
the use of socket options for multihoming shim layer.</t>
<section title="Implicit Forking">
<t>Socket options defined in <xref
target="sec-shim-socket-options"/> may cause conflicting
situation when the target context is shared by multiple
applications. In such case, socket handler and the
multihoming shim layer should react as follows; socket handler
should inform the shim layer that context forking is required.
In SHIM6, when a context is forked, an unique identifier
called Forked Instance Identifier (FII) is assigned to the
newly forked context. The forked context is then exclusively
associated with the socket through which non-default
preference value was specified. The forked context is
maintained by the multihoming shim layer during the lifetime
of associated socket instance. When the socket is closed, the
multihoming shim layer SHOULD delete associated context. In
this way, garbage collection can be carried out to cleanup
unused forked contexts. Upon garbage collection, every forked
context SHOULD be checked if there is no socket (process)
associated with the context. If there is none, the forked
context should be deleted. When a forked context is torn
down, SHIM6 should notify the peer about the deletion of
forked context.</t>
<t>As opposed to socket options, context forking MUST NOT be
triggered by any use of ancillary data that is specific to
multihoming shim as defined in <xref
target="sec-access-to-locinfo"/>.</t>
</section>
<!--
<section title="Explicit Forking">
<t>There is another approach to support context forking by
multihoming shim API. In this approach, it is assumed that
shim-aware application can make distinction of each shim
context that has the same EID pair and specify which context
to be used for its communication.</t>
<t>Socket option SHIM_TBD1 can be used by application to
request the multihoming shim layer to fork a context. If the
context is successfully forked by the shim, the FII assigned
for the forked context is returned to the application. The
application can request the multihoming shim layer to apply
specific context to its communication either by socket option
or ancillary data specifying SHIM_TBD2 socket option along
with FII value.</t>
<t>NOTE: If we decide to go with explicit forking, we probably
need to define comprehensive set of socket options that allow
application to specify adaptation of forked context to the
flow over a given socket. It seems to me that there are lots
of things to do to support explicit model.</t>
</section>
-->
</section>
<section title="Discussion" anchor="sec-discussion" toc="include">
<t>In this section, open issues are introduced.</t>
<!--
<section title="Issues with a Context Shared by Applications">
<t>A context is by definition, system-wide. This means that a
context could be shared by applications whose communications
are using the same EID pair.</t>
<t>When a context is shared by applications, there may be some
problems when the shim layer needs to handle feedbacks from
the multiple applications. As mentioned in <xref
target="sec-access-to-locinfo"/>, an application may provide
the shim layer feedback about timeout values from its own
settings. This implies that there is potentially a race
condition at the shim layer.</t>
<t>First of all, the socket options must be used with a proper
privilege. Feedback from the application which is run under a
root privilege must always override the feedback provided by
application which is run under normal user privilege.</t>
<t>For other cases, one could rely on a kind of heuristics of
the configuration. For instance, prioritizing feedback with
higher demand (e.g. timeout value 300 seconds are more
demanding then timeout value 600 seconds) may make sense in
some cases. However, it is still an open issue what kind of
timer value could be handled in this way.</t>
<t>Further discussions are needed how the shim layer can
accommodate feedbacks from multiple applications within a same
context.</t>
</section>
-->
<!--
<section title="Issues with Shim Unaware Application">
<t>In multihomed environment where either of the peers or both
of the peers have multiple locators, there are some issues
with shim unaware application which uses legacy socket
API.</t>
<section title="Initial Contact with Multiple Locator Pairs">
<t>In a connection oriented communication, the connect()
system call is used to make the initial contact to the peer,
which typically requires IP address and port number to
specify the endpoint. Hence, name-to-address resolution
should be performed prior to connect(). The application
needs to resolve the FQDN of the peer to an IP address by
any available name-to-address conversion method.</t>
<t>In typical case, the application receives information
from the resolver. If the application ends up with
receiving multiple IP addresses to reach the peer, it should
iterate through each destination address one-by-one. It
should be noted that the host may also have multiple source
addresses.</t>
<t>The different resulting address pairs may have different
reachability status so, in order to find a working address
pair, it may be required to explore all the available
address pairs (as opposed to explore all available
destination addresses).</t>
<t>In normal case, the application issues a connect() by
specifying the resolved IP address of the peer. If the
connect() fails, it iterates through the available IP
addresses one by one sequentially until working pair is
found. Another approach is to initiate concurrent connect()
with every locator of the peer. connect() can also be
called in a sequence which would probably require more time
to find the working pair.</t>
<t>There is a case where involvement of the shim layer is
expected for handling initial contact. In such case,
behavior of the shim layer will depend on presence of the
required context. This case occurs when there exists a
context for the EID specified in connect(), the initial
contact can be made in accordance with the context
information. Otherwise, the shim layer should invoke
context establishment with the peer EID specified in the
argument for connect().</t>
<t>Additional efforts would be required in a case where the
peer cannot be reachable through the EID (for example, EID
is non-routable or non-IP reachable) but it can be reached
through alternative locator. In particular, the shim layer
should somehow discover the alternate locator for the EID to
establish context. <xref target="I-D.nordmark-shim6-esd"/>
addresses the possible approach to perform reverse DNS
lookup from EID to FQDN, then perform forward lookup again
to find the full-set of locators and EID.</t>
<t>In HIP, resolving HITs to IP addresses using DNS is not
feasible because HITs do not contain any hierarchical
information. To mitigate this problem, there are a few
alternatives. Firstly, resolver library on end-host can be
modified to provide HIT-to-IP mappings for HIP software
module. Secondly, a distributed hash table (DHT) service
can be used for storing and looking up the mappings because
it supports non-hierarchical identifiers, such as HITs <xref
target="RFC4423"/>. Thirdly, it is possible to
use IP addresses in legacy applications as described in
<xref target="I-D.henderson-hip-applications"/>.</t>
</section>
</section>
-->
<section title="Naming at Socket Layer">
<t>The getsockname() and getpeername() system calls are used
to obtain the 'name' of an endpoint which is actually a pair
of IP address and port number assigned to a given socket.
getsockname() is used when an application wants to obtain the
local IP address and port number assigned for a given socket
instance. getpeername() is used when an application obtains
the remote IP address and port number.</t>
<t>The above is based on a traditional system model of the
sockets API where an IP address is expected to play both the
role of identifier and the role of locator.</t>
<t>In a system model where a shim layer exists inside the IP
layer, both getsockname() and getpeername() deal with
identifiers, namely EIDs. In this sense, the shim layer
serves to (1) hide locators and (2) provide access to the
identifier for the application over the legacy socket
APIs.</t>
</section>
<section title="Additional Requirements from Applications">
<t>At the moment, it is not certain if following requirements
are common in all the multihomed environments (SHIM6 and HIP).
These are mainly identified during discussions made on SHIM6
WG mailing list.
<list style="symbols">
<t>The application should be able to set preferences for the
locators, local and remote ones, and also to the preferences
of the local locators that will be passed to the peer.</t>
</list>
</t>
</section>
<section title="Issues of Header Conversion among Different Address Family">
<t>The shim layer performs identifier/locator adaptation.
Therefore, in some case, the whole IP header can be replaced
with new IP header of a different address family
(e.g. conversion from IPv4 to IPv6 or vice versa). Hence,
there is an issue how to make the conversion with minimum
impact. Note that this issue is common in other protocol
conversion such as SIIT<xref target="RFC2765"/>.</t>
<t>As addressed in SIIT specification, some of the features
(IPv6 routing headers, hop-by-hop extension headers, or
destination headers) from IPv6 are not convertible to IPv4.
In addition, notion of source routing is not exactly the same
in IPv4 and IPv6. Hence, there is certain limitation in
protocol conversion between IPv4 and IPv6.</t>
<t>The question is how should the shim layer behave when it is
face with limitation problem of protocol conversion. Should we
introduce new error something like ENOSUITABLELOCATOR ?</t>
</section>
<section title="Handling of Unknown Locator Provided by Application">
<t>There might be a case where application provides the shim
layer new locator with the SHIM_LOC_*_PREF socket options or
SHIM_LOC_*_SEND ancillary data. Then there is a question how
should the shim layer treat the new locator informed by the
application.</t>
<t>In principle, locator information are exchanged by the shim
protocol. However, there might be a case where application
acquires information about the locator and prefers to use it
for its communication.</t>
</section>
</section>
<!--
================================================================
Changes
================================================================
-->
<section title="Changes" toc="default">
<section title="Changes from version 00 to version 01">
<t>The followings are changes from version 00 to version 01:
<list style="symbols">
<t>Define shim_locator{} data type which is a placeholder for
locator.</t>
<t>Define shim_pathexplore{} data type in which a set of
REAP parameters are stored.</t>
<t>Remove descriptions about "stickiness" of socket options.</t>
<t>Deprecate SHIM_IF_RECV and SHIM_IF_SEND socket options.</t>
<t>Give default value and how to disable given socket
option.</t>
</list>
</t>
</section>
<section title="Changes from version 01 to version 02">
<t>The followings are changes from version 01 to version 02:
<list style="symbols">
<t>Add section describing context forking.</t>
<t>Rephrase conclusion section.</t>
<t>Separate normative references from informative
references.</t>
<t>Remove texts from discussion section that are not
relevant to the contents of the document.</t>
<t>Add section describing change history (this section).</t>
</list>
</t>
</section>
<section title="Changes from version 02 to version 03">
<t>The followings are changes from version 02 to version 03:
<list style="symbols">
<t>Add an Appendix section describing the issue of context
forking.</t>
</list>
</t>
</section>
<section title="Changes from version 03 to version 04">
<t>The followings are changes from version 03 to version 04:
<list style="symbols">
<t>Updated reference.</t>
<t>Correct typo and grammatical errors.</t>
</list>
</t>
</section>
<section title="Changes from version 04 to version 05">
<t>The followings are changes from version 04 to version 05:
<list style="symbols">
<t>Added definition of SHIM_FEEDBACK ancillary data.</t>
<t>Added an example of code using the SHIM_LOCLIST_LOCAL</t>
<t>Added SHIM_LOC_LOCAL_SEND and SHIM_LOC_PEER_SEND socket
options.</t>
</list>
</t>
</section>
<section title="Changes from version 05 to version 06">
<t>The followings are changes from version 04 to version 05:
<list style="symbols">
<t>Updated references.</t>
</list>
</t>
</section>
<section title="Changes from version 06 to version 07">
<t>The followings are changes from version 06 to version 07:
<list style="symbols">
<t>Resolved editorial issues.</t>
</list>
</t>
</section>
</section>
<!--
================================================================
IANA Consideration
================================================================
-->
<section title="IANA Considerations" toc="default">
<t>This document contains no IANA consideration.</t>
</section>
<!--
================================================================
Security Consideration
================================================================
-->
<section title="Security Considerations" toc="default">
<t>This document does not specify any security mechanism for the
shim layer. Fundamentally, the shim layer has a potential to
impose security threats, as it changes the source and/or
destination IP addresses of the IP packet being sent or
received. Therefore, the basic assumption is that the security
mechanism defined in each protocol of the shim layer is strictly
applied.</t>
<!-- should we mention about privilege ? -->
<!-- marcelo: perhaps we should talk about what happens when the
app includes a new locator in the locator set... whether
this would be possible or not... -->
<!-- shinta: i've put the above issue in discussion section -->
</section>
<!--
================================================================
Conclusion
================================================================
-->
<section title="Conclusion" toc="default">
<t>In this document, the Application Program Interface (API) for
multihoming shim layer is specified. The sockets API allows
applications to have additional control of the locator
management and interface to the REAP mechanism inside the
multihoming shim layer.</t>
<t>Socket options for multihoming shim layer can be used by
getsockopt() and/or setsockopt() system calls. Besides,
applications can use some ancillary data that are specific to
multihoming shim layer to get locator from received packet or to
set locator for outgoing packet.</t>
<t>From an architectural point of view, the sockets API provides
extends the existing sockets API framework in the face of
ID/Locator separation. With regard to API that relate to IP
address management, it is assured that existing sockets API
continue to work above the shim layer dealing with identifiers,
while multihoming shim API deals with locators.</t>
</section>
<!--
================================================================
Acknowledgment
================================================================
-->
<section title ="Acknowledgments" toc="include">
<t>Authors would like to thank Jari Arkko who participated in
the discussion that lead to the first version of this document,
and Tatuya Jinmei who thoroughly reviewed the early version of
this draft and provided detailed comments on sockets API related
issues. Thomas Henderson provided valuable comments especially
from HIP perspectives.</t>
</section>
</middle>
<back>
<!--
================================================================
References
================================================================
-->
<references title="Normative References">
<reference anchor='I-D.ietf-shim6-proto'>
<front>
<title>Level 3 multihoming shim protocol</title>
<author initials='M' surname='Bagnulo' fullname='Marcelo Bagnulo'>
<organization />
</author>
<author initials='E' surname='Nordmark' fullname='Erik Nordmark'>
<organization />
</author>
<date month='February' year='2008' />
</front>
<seriesInfo name='Internet-Draft' value='draft-ietf-shim6-proto-10' />
<format type='TXT'
target='http://www.ietf.org/internet-drafts/draft-ietf-shim6-proto-10.txt' />
</reference>
<reference anchor='RFC4423'>
<front>
<title>Host Identity Protocol (HIP) Architecture</title>
<author initials='R.' surname='Moskowitz' fullname='R. Moskowitz'>
<organization /></author>
<author initials='P.' surname='Nikander' fullname='P. Nikander'>
<organization /></author>
<date year='2006' month='May' />
</front>
<seriesInfo name='RFC' value='4423' />
<format type='TXT' octets='61049' target='ftp://ftp.isi.edu/in-notes/rfc4423.txt' />
</reference>
<!--
<reference anchor='RFC3493'>
<front>
<title>Basic Socket Interface Extensions for IPv6</title>
<author initials='R.' surname='Gilligan' fullname='R. Gilligan'>
<organization /></author>
<author initials='S.' surname='Thomson' fullname='S. Thomson'>
<organization /></author>
<author initials='J.' surname='Bound' fullname='J. Bound'>
<organization /></author>
<author initials='J.' surname='McCann' fullname='J. McCann'>
<organization /></author>
<author initials='W.' surname='Stevens' fullname='W. Stevens'>
<organization /></author>
<date year='2003' month='February' /></front>
<seriesInfo name='RFC' value='3493' />
<format type='TXT' octets='82570'
target='ftp://ftp.isi.edu/in-notes/rfc3493.txt' />
</reference>
-->
<reference anchor='RFC3542'>
<front>
<title>Advanced Sockets Application Program Interface (API)
for IPv6</title>
<author initials='W.' surname='Stevens' fullname='W. Stevens'>
<organization /></author>
<author initials='M.' surname='Thomas' fullname='M. Thomas'>
<organization /></author>
<author initials='E.' surname='Nordmark' fullname='E. Nordmark'>
<organization /></author>
<author initials='T.' surname='Jinmei' fullname='T. Jinmei'>
<organization /></author>
<date year='2003' month='May' /></front>
<seriesInfo name='RFC' value='3542' />
<format type='TXT' octets='173028'
target='ftp://ftp.isi.edu/in-notes/rfc3542.txt' />
</reference>
<reference anchor='I-D.ietf-shim6-failure-detection'>
<front>
<title>Failure Detection and Locator Pair Exploration
Protocol for IPv6 Multihoming</title>
<author initials='J' surname='Arkko' fullname='Jari Arkko'>
<organization />
</author>
<author initials='I' surname='Beijnum' fullname='Iljitsch van Beijnum'>
<organization />
</author>
<date month='June' day='24' year='2008' />
</front>
<seriesInfo name='Internet-Draft' value='draft-ietf-shim6-failure-detection-13' />
<format type='TXT'
target='http://www.ietf.org/internet-drafts/draft-ietf-shim6-failure-detection-13.txt' />
<format type='PDF'
target='http://www.ietf.org/internet-drafts/draft-ietf-shim6-failure-detection-13.pdf' />
</reference>
<reference anchor='POSIX'>
<front>
<title>IEEE Std. 1003.1-2001 Standard for Information
Technology -- Portable Operating System Interface
(POSIX). Open group Technical Standard: Base Specifications,
Issue 6, http://www.opengroup.org/austin</title>
<author initials='' surname=''
fullname='IEEE Standard'>
<organization />
</author>
<date month='December' day='1' year='2001' />
</front>
<format type='TXT'
target='http://www.opengroup.org/austin' />
</reference>
<!--
<reference anchor='I-D.henderson-hip-applications'>
<front>
<title>Using HIP with Legacy Applications</title>
<author initials='T' surname='Henderson' fullname='Tom Henderson'>
<organization />
</author>
<author initials='P' surname='Nikander' fullname='Pekka Nikander'>
<organization />
</author>
<date month='May' day='17' year='2006' />
</front>
<seriesInfo name='Internet-Draft'
value='draft-henderson-hip-applications-03' />
<format type='TXT'
target='http://www.ietf.org/internet-drafts/draft-henderson-hip-applications-03.txt' />
</reference>
-->
</references>
<references title="Informative References">
<reference anchor='I-D.ietf-shim6-app-refer'>
<front>
<title>Shim6 Application Referral Issues</title>
<author initials='E' surname='Nordmark' fullname='Erik Nordmark'>
<organization />
</author>
<date month='July' day='5' year='2005' />
</front>
<seriesInfo name='Internet-Draft' value='draft-ietf-shim6-app-refer-00' />
<format type='TXT'
target='http://www.ietf.org/internet-drafts/draft-ietf-shim6-app-refer-00.txt' />
</reference>
<!--
<reference anchor='I-D.nordmark-shim6-esd'>
<front>
<title>Extended Shim6 Design for ID/loc split and Traffic
Engineering</title>
<author initials='E' surname='Nordmark' fullname='Erik Nordmark'>
<organization />
</author>
<date month='February' day='26' year='2006'/>
</front>
<seriesInfo name='Internet-Draft' value='draft-nordmark-shim6-esd-00' />
<format type='TXT'
target='http://www.ietf.org/internet-drafts/draft-nordmark-shim6-esd-00.txt' />
</reference>
-->
<reference anchor='RFC3972'>
<front>
<title>Cryptographically Generated Addresses (CGA)</title>
<author initials='T.' surname='Aura' fullname='T. Aura'>
<organization /></author>
<date year='2005' day='11' month='March' /></front>
<seriesInfo name='RFC' value='3972' />
<format type='TXT' octets='51473'
target='ftp://ftp.isi.edu/in-notes/rfc3972.txt' />
</reference>
<reference anchor='I-D.ietf-shim6-hba'>
<front>
<title>Hash Based Addresses (HBA)</title>
<author initials='M' surname='Bagnulo' fullname='Marcelo Bagnulo'>
<organization />
</author>
<date month='December' day='22' year='2007' />
</front>
<seriesInfo name='Internet-Draft' value='draft-ietf-shim6-hba-05' />
<format type='TXT'
target='http://www.ietf.org/internet-drafts/draft-ietf-shim6-hba-05.txt' />
</reference>
<reference anchor='RFC2765'>
<front>
<title abbrev='SIIT'>Stateless IP/ICMP Translation Algorithm (SIIT)</title>
<author initials='E.' surname='Nordmark' fullname='Erik Nordmark'>
<organization /></author>
<date year='2000' month='February' />
</front>
<seriesInfo name='RFC' value='2765' />
<format type='TXT' octets='59465' target='ftp://ftp.isi.edu/in-notes/rfc2765.txt' />
</reference>
<reference anchor="RFC4291">
<front>
<title>IP Version 6 Addressing Architecture</title>
<author initials='R.' surname='Hinden' fullname='R. Hinden'>
<organization /></author>
<author initials='S.' surname='Deering' fullname='S. Deering'>
<organization /></author>
<date year='2006' month='February' />
</front>
<seriesInfo name='RFC' value='4291' />
<format type='TXT' octets='52897'
target='ftp://ftp.isi.edu/in-notes/rfc4291.txt' />
</reference>
</references>
<section title="Context Forking">
<t>In this section, an issue concerning context forking and its
relation to the multihoming shim API are discussed.</t>
<t>SHIM6 supports a notion of context forking. A peer may
decide to fork a context for certain reason (e.g. upper layer
protocol prefers to use different locator pair than the one
defined in available context). The procedure of forking context
is done similar to the normal context establishment, performing
the 4-way message exchange. A peer who has decided to fork a
context initiates the context establishment. Hereafter, we call
this peer initiator.</t>
<t>Once the forked context is established between the peers, on
the initiator side, it is possible to apply forked context to
the packet flow since the system maintains an association
between the forked context and the socket owned by the
application that has requested the context forking. How this
association is maintained is implementation specific issue.
However, on the responder side, there is a question on how the
outbound packet can be multiplexed by the shim layer. Since
there are more than one SHIM6 contexts that match with the ULID
pair of the packet flow. There is a need to differentiate
packet flows not only by the ULID pairs but some other
information and associate a given packet flow with specific
context.</t>
<t><xref target="fig-context-forking"/> gives an example of a
scenario where two communicating peers fork a context.
Initially, there has been a single transaction between the
peers, by the application 1 (App1). Accordingly, another
transaction is started, by application 2 (App2). Both of the
transactions are made based the same ULID pair. The first
context pair (Ctx1) is established for the transaction of App1.
Given the requests from App2, the shim layer on Peer 1 decides
to fork a context. Accordingly, a forked context (Ctx2) is
established between the peers, which should be exclusively
applied to the transaction of App2. Ideally, multiplexing and
demultiplexing of packet flows that relate to App1 and App2
should be done as illustrated in <xref
target="fig-context-forking"/>. However, as mentioned earlier,
the responder needs to multiplex outbound flows of App1 and App2
somehow. Note that if a context forking occurs on the initiator
side, a context forking needs to occur also on the responder
side.
<figure anchor="fig-context-forking" title="context forking">
<artwork><![CDATA[
Peer 1 Peer 2
(initiator) (responder)
+----+ +----+ +----+ +----+
|App1| |App2| |App1| |App2|
+----+ +----+ +----+ +----+
|^ |^ ^| ^|
v| v| |v |v
-----S1-------------S2----- -----S1-------------S2-----
|| || || ||
|| || || ||
Ctx1 Ctx2 Ctx1 Ctx2
ULID:<A1,B1> ULID:<A1,B1> ULID:<B1,A1> ULID:<B1,A1>
Loc: <A1,B2> Loc: <A1,B3> Loc: <B2,A1> Loc: <B3,A1>
FII: 0 FII: 100 FII: 0 FII: 100
|^ |^ ^| ^|
|| || || ||
|| || || ||
\..............||........................../| ||
\.............||.........................../ ||
|| ||
\|........................................./|
\........................................../
]]></artwork>
</figure>
</t>
<t>To overcome the problem mentioned above, there are some
solutions.</t>
<t>One viable approach is to let the system implicitly maintain
an association between the socket and the associated context by
keeping the record of inbound packet processing. That is, the
system stores the information about the context on which the
inbound packet flow was demultiplexed. The information
comprises the ULID pair and FII of the context and is stored in
the socket instance. Later, the system can use the information
to identify the associated context in outbound packet
processing. This approach should be feasible as far as there is
bi-directional user traffic.</t>
<t>Another viable approach is to extend SHIM6 protocol by adding
capability of exchanging additional information to identify the
packet flow from others which needs to be handled by a newly
forked context. The information exchange can be done during the
context establishment. The initiator appends 5 tuple of the
packet flow to be handled by the newly forked context. Note
that the additional information provided by the 5 tuple are
source and destination port numbers and upper layer protocol.
The information is later used by the shim layer to multiplex the
outbound packet flow on the responder side.</t>
<t>The socket options for multihoming shim can be used by the
application to trigger the context forking in implicit manner.
The peer becomes an initiator in the establishment of the forked
context. Once the forked context is established between the
peers, application on each end can influence the preference on
context by utilizing the multihoming shim API.</t>
</section>
</back>
</rfc>
| PAFTECH AB 2003-2026 | 2026-04-22 22:47:39 |