draft-ietf-dhc-dual-stack-01.txt   draft-ietf-dhc-dual-stack-02.txt 
Dynamic Host Congiguration T. Chown Dynamic Host Congiguration T. Chown
Internet-Draft University of Southampton Internet-Draft University of Southampton
Expires: January 17, 2005 S. Venaas Expires: April 25, 2005 S. Venaas
UNINETT UNINETT
C. Strauf C. Strauf
Technical University of Clausthal Technical University of Clausthal
July 19, 2004 October 25, 2004
DHCP: IPv4 and IPv6 Dual-Stack Issues DHCP: IPv4 and IPv6 Dual-Stack Issues
draft-ietf-dhc-dual-stack-01 draft-ietf-dhc-dual-stack-02
Status of this Memo Status of this Memo
By submitting this Internet-Draft, I certify that any applicable By submitting this Internet-Draft, I certify that any applicable
patent or other IPR claims of which I am aware have been disclosed, patent or other IPR claims of which I am aware have been disclosed,
and any of which I become aware will be disclosed, in accordance with and any of which I become aware will be disclosed, in accordance with
RFC 3668. RFC 3668.
Internet-Drafts are working documents of the Internet Engineering Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that Task Force (IETF), its areas, and its working groups. Note that
skipping to change at page 1, line 36 skipping to change at page 1, line 36
and may be updated, replaced, or obsoleted by other documents at any and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress." material or to cite them other than as "work in progress."
The list of current Internet-Drafts can be accessed at The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt. http://www.ietf.org/ietf/1id-abstracts.txt.
The list of Internet-Draft Shadow Directories can be accessed at The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html. http://www.ietf.org/shadow.html.
This Internet-Draft will expire on January 17, 2005. This Internet-Draft will expire on April 25, 2005.
Copyright Notice Copyright Notice
Copyright (C) The Internet Society (2004). All Rights Reserved. Copyright (C) The Internet Society (2004). All Rights Reserved.
Abstract Abstract
A node may have support for communications using IPv4 and/or IPv6 A node may have support for communications using IPv4 and/or IPv6
protocols. Such a node may wish to obtain IPv4 and/or IPv6 protocols. Such a node may wish to obtain IPv4 and/or IPv6
configuration settings via the Dynamic Host Configuration Protocol configuration settings via the Dynamic Host Configuration Protocol
(DHCP). The original version of DHCP [1] designed for IPv4 has now (DHCP). The original version of DHCP [1] designed for IPv4 has now
been complemented by a new DHCPv6 [4] for IPv6. This document been complemented by a new DHCPv6 [4] for IPv6. This document
describes issues identified with dual IP version DHCP interactions, describes issues identified with dual IP version DHCP interactions,
the most important aspect of which is how to handle potential the most important aspect of which is how to handle potential
problems in clients processing configuration information received problems in clients processing configuration information received
from DHCPv4 and DHCPv6 servers. from both DHCPv4 and DHCPv6 servers. The document makes a
recommendation on the general strategy on how best to handle such
issues, and identifies future work to be undertaken.
Table of Contents Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3
2. Configuration scenarios . . . . . . . . . . . . . . . . . . . 3 2. Configuration scenarios . . . . . . . . . . . . . . . . . . . 3
3. Dual-stack issues . . . . . . . . . . . . . . . . . . . . . . 4 3. Dual-stack issues . . . . . . . . . . . . . . . . . . . . . . 4
3.1 Handling multiple responses . . . . . . . . . . . . . . . 4 3.1 Handling multiple responses . . . . . . . . . . . . . . . 4
3.2 Different administrative management . . . . . . . . . . . 4 3.2 Different administrative management . . . . . . . . . . . 5
3.3 Multiple interfaces . . . . . . . . . . . . . . . . . . . 5 3.3 Multiple interfaces . . . . . . . . . . . . . . . . . . . 5
3.4 DNS load balancing . . . . . . . . . . . . . . . . . . . . 5 3.4 DNS load balancing . . . . . . . . . . . . . . . . . . . . 5
3.5 DNS search path issues . . . . . . . . . . . . . . . . . . 5 3.5 DNS search path issues . . . . . . . . . . . . . . . . . . 5
3.6 Protocol startup sequence . . . . . . . . . . . . . . . . 5 3.6 Protocol startup sequence . . . . . . . . . . . . . . . . 5
3.7 DHCP option variations . . . . . . . . . . . . . . . . . . 5 3.7 DHCP option variations . . . . . . . . . . . . . . . . . . 6
3.8 Security issues . . . . . . . . . . . . . . . . . . . . . 6 3.8 Security issues . . . . . . . . . . . . . . . . . . . . . 6
4. Potential solutions . . . . . . . . . . . . . . . . . . . . . 6 4. Potential solutions . . . . . . . . . . . . . . . . . . . . . 6
4.1 Separate DHCP servers . . . . . . . . . . . . . . . . . . 6 4.1 Separate DHCP servers . . . . . . . . . . . . . . . . . . 6
4.2 Single DHCPv6 server . . . . . . . . . . . . . . . . . . . 7 4.2 Single DHCPv6 server . . . . . . . . . . . . . . . . . . . 7
4.3 Administrative and other areas . . . . . . . . . . . . . . 8 4.3 Optimising for failure with lists of addresses . . . . . . 8
4.4 Administrative and other areas . . . . . . . . . . . . . . 9
5. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 5. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
6. Security Considerations . . . . . . . . . . . . . . . . . . . 10 6. Security Considerations . . . . . . . . . . . . . . . . . . . 11
7. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 10 7. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 11
8. Informative References . . . . . . . . . . . . . . . . . . . . 10 8. Informative References . . . . . . . . . . . . . . . . . . . . 11
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . 11 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . 12
Intellectual Property and Copyright Statements . . . . . . . . 12 Intellectual Property and Copyright Statements . . . . . . . . 13
1. Introduction 1. Introduction
The original specification of the Dynamic Host Configuration Protocol The original specification of the Dynamic Host Configuration Protocol
(DHCP) was made with only IPv4 in mind. That specification has been (DHCP) was made with only IPv4 in mind. That specification has been
subsequently revised, up to the latest version of DHCP [1]. With the subsequently revised, up to the latest version of DHCP [1]. With the
arrival of IPv6, a new DHCP specification for IPv6 has been designed, arrival of IPv6, a new DHCP specification for IPv6 has been designed,
and published as DHCPv6 [4]. and published as DHCPv6 [4].
These protocols allow nodes to communicate via IPv4 or IPv6 to These protocols allow nodes to communicate via IPv4 or IPv6 to
retrieve configuration settings for operation in a managed retrieve configuration settings for operation in a managed
environment. While an IPv6 node may acquire address-related environment. While an IPv6 node may acquire address-related
configuration settings via IPv6 stateless address autoconfiguration configuration settings via IPv6 stateless address autoconfiguration
[2], such a node may wish to use stateless DHCPv6 [5] for other [2], such a node may wish to use stateless DHCPv6 [5] for other
administratively configured options (e.g. DNS, NTP). administratively configured options, such as DNS or NTP.
In early IPv6 deployments, a dual-stack mode of operation is In early IPv6 deployments, a dual-stack mode of operation is
typically used. There will thus be nodes that require both IPv4 and typically used. There will thus be nodes that require both IPv4 and
IPv6 configuration settings. This document discusses issues with IPv6 configuration settings. This document discusses issues with
obtaining such settings in a dual-stack environment. obtaining such settings in a dual-stack environment.
In this document, we refer to a "DHCP server" as a server In this document, we refer to a "DHCP server" as a server
implementing the original DHCP [1], and a "DHCPv6 server" as a server implementing the original DHCP [1], and a "DHCPv6 server" as a server
implementing DHCPv6 [4] or its stateless subset [5]. implementing DHCPv6 [4] or its stateless subset [5].
skipping to change at page 3, line 52 skipping to change at page 3, line 52
o NTP server o NTP server
o DNS server o DNS server
o NIS server o NIS server
o DNS search path o DNS search path
While the format of address settings will be IP-specific, the node While the format of address settings will be IP-specific, the node
may equally well acquire IPv4 or IPv6 addresses for some settings, may equally well acquire IPv4 or IPv6 addresses for some settings,
e.g. for DNS or NTP, if those services are available via IPv4 or such as for DNS or NTP, if those services are available via IPv4 or
IPv6 transport. Currently, a DHCP server returns IPv4 data, while a IPv6 transport. Currently, a DHCP server returns IPv4 data, while a
DHCPv6 server returns IPv6 data. DHCPv6 server returns IPv6 data.
It is worth noting that in an IPv4 environment, with a DHCP server, It is worth noting that in an IPv4 environment, with a DHCP server,
the choice of whether to use DHCP is made by the node. In an IPv6 the choice of whether to use DHCP is made by the node. In an IPv6
environment, the use of the managed and other bits in the Router environment, the use of the managed and other bits in the Router
Advertisement can offer a hint to the node whether or not to use full Advertisement can offer a hint to the node whether or not to use full
DHCPv6 or its stateless variant. It is perhaps not clear whether a DHCPv6 or its stateless variant. It is perhaps not clear whether a
dual-stack node should do DHCP for IPv4 if Managed and OtherConfig dual-stack node should do DHCP for IPv4 if Managed and OtherConfig
flags in the Router Advertisement are both off; it seems most flags in the Router Advertisement are both off; it seems most
skipping to change at page 4, line 32 skipping to change at page 4, line 32
It has been noted from comments that the first four, and possibly It has been noted from comments that the first four, and possibly
five, subsections here may also be viewed as multihoming issues. five, subsections here may also be viewed as multihoming issues.
3.1 Handling multiple responses 3.1 Handling multiple responses
The general question is how to handle configuration information that The general question is how to handle configuration information that
may be gathered from multiple sources. Where those sources are DHCP may be gathered from multiple sources. Where those sources are DHCP
and DHCPv6 servers (which may be two physical nodes or two servers and DHCPv6 servers (which may be two physical nodes or two servers
running on the same node) the client node needs to know whether to running on the same node) the client node needs to know whether to
use the most recent data, or whether to perform some merger or union use the most recent data, or whether to perform some merger or union
of the responses by certain rules. A node may choose to ask a DHCPv6 of the responses by certain rules. A method for merging lists of
server and only use a DHCP server if no response is received. addresses, for options that carry such information, may also be
required. A node may choose to ask a DHCPv6 server and only use a
DHCP server if no response is received.
Merging is possible, but is likely to be complex. There could be Merging is possible, but is likely to be complex. There could be
some priority, so that if both DHCP and DHCPv6 servers offer a value, some priority, so that if both DHCP and DHCPv6 servers offer a value,
only one is used. Or the node could choose to store and use both, in only one is used. Or the node could choose to store and use both, in
some order of its choosing. some order of its choosing. Merging issues are discussed further
later in this document.
A node may also obtain information from other sources, e.g. a manual A node may also obtain information from other sources, such as a
configuration file (e.g. /etc/resolv.conf for DNS data on many Unix manual configuration file (for example /etc/resolv.conf for DNS data
systems). A node configured manually to use an IPv6 DNS server via on many UNIX systems). A node configured manually to use an IPv6 DNS
such manual configuration may lose that configuration if it then uses server may lose that configuration if it is in a dual-stack
DHCP to obtain IPv4 settings if in a dual-stack environment; that environment and uses DHCP to obtain IPv4 settings; the new IPv4
IPv4 configuration may then overwrite the manual IPv6 DNS setting settings from the DHCP response may then overwrite the manual IPv6
with new IPv4 settings from the DHCP response. DNS setting.
3.2 Different administrative management 3.2 Different administrative management
In some deployments, the IPv4 and IPv6 services may not be In some deployments, the IPv4 and IPv6 services may not be
administered by the same organisation or people, e.g. in a community administered by the same organisation or people, such as in a
wireless environment. This poses problems for consistency of data community wireless environment. This poses problems for consistency
offered by either DHCP version. of data offered by either DHCP version.
3.3 Multiple interfaces 3.3 Multiple interfaces
A node may have multiple interfaces and run IPv4 and IPv6 on A node may have multiple interfaces and run IPv4 and IPv6 on
different interfaces. A question then is whether the settings are different interfaces. A question then is whether the settings are
per interface or per node? DHCPv6 introduces the idea of a DHCP per interface or per node? DHCPv6 introduces the idea of a DHCP
Unique Identifier (DUID) which does not yet exist for DHCP; some Unique Identifier (DUID) which does not yet exist for DHCP; some
effort is being made to retrofit the concept to DHCPv4 [6]. effort is being made to retrofit the concept to DHCPv4 [6].
Per interface settings can be complex because a client node needs to Per interface settings can be complex because a client node needs to
know from which interface system settings like NTP server came from. know from which interface system settings like NTP server came from.
And it may not be apparent which setting should be used, if e.g. an And it may not be apparent which setting should be used, if e.g. an
NTP server option is received on multiple interfaces, potentially NTP server option is received on multiple interfaces, potentially
over different protocols. over different protocols.
3.4 DNS load balancing 3.4 DNS load balancing
In some cases it is preferable to list DNS server information in an In some cases it is preferable to list DNS server information in an
ordered way per node for load balancing, giving different responses ordered way per node for load balancing, giving different responses
to different clients. Responses from different DHCP and DHCPv6 to different clients. Responses from different DHCP and DHCPv6
servers may make such configuration problematic. servers may make such configuration problematic, if the knowledge of
the load balancing is not available to both servers.
3.5 DNS search path issues 3.5 DNS search path issues
The DNS search path may vary for administrative reasons. For The DNS search path may vary for administrative reasons. For
example, a site under the domain foo.com chooses to place an early example, a site under the domain foo.com chooses to place an early
IPv6 deployment under the subdomain ipv6.foo.com, until it is IPv6 deployment under the subdomain ipv6.foo.com, until it is
confident of offering a full dual-stack service under its main confident of offering a full dual-stack service under its main
domain. The subtlety here is that the DNS search path then affects domain. The subtlety here is that the DNS search path then affects
choice of protocol used, e.g. IPv6 for nodes in ipv6.foo.com. choice of protocol used, such as IPv6 for nodes in ipv6.foo.com.
3.6 Protocol startup sequence 3.6 Protocol startup sequence
In the dual-stack environment, one needs to consider what happens if, In the dual-stack environment, one needs to consider what happens if,
for example, the IPv6 interface (transport) is started after DHCPv4 for example, the IPv6 interface (transport) is started after DHCPv4
was used to configure the client. Should the client then simply was used to configure the client. Should the client then simply
discard the current IPv4 information, or merge it with a subsequent discard the current IPv4 information, or merge it with a subsequent
IPv6 response? IPv6 response?
3.7 DHCP option variations 3.7 DHCP option variations
Some options in DHCP are not available in DHCPv6 and vice-versa. Some options in DHCP are not available in DHCPv6 and vice-versa.
Some IP-version limitations naturally apply, e.g. only IPv6 Some IP-version limitations naturally apply, such as only IPv6
addresses can be in an IPv6 NTP option. The DHCP and DHCPv6 option addresses can be in an IPv6 NTP option. The DHCP and DHCPv6 option
numbers may be different. numbers may be different.
There may be some sites that would choose to use IPv4-mapped Some sites may choose to use IPv4-mapped addresses in DHCPv6-based
addresses in DHCPv6-based options. The merits and drawbacks of such options. The merits and drawbacks of such an approach need
an approach need discussion. discussion.
A site administrator may wish to configure all their dual-stack nodes A site administrator may wish to configure all their dual-stack nodes
with (say) two NTP servers, one of which has an IPv4 address, the with (say) two NTP servers, one of which has an IPv4 address, the
other an IPv6 address. In this case it may be desirable for an NTP other an IPv6 address. In this case it may be desirable for an NTP
option to carry a list of addresses, where some may be IPv4 and some option to carry a list of addresses, where some may be IPv4 and some
may be IPv6. In general one could consider having DHCPv6 options may be IPv6. In general one could consider having DHCPv6 options
that can carry mix of IPv4 and IPv6 addresses. that can carry a mix of IPv4 and IPv6 addresses.
3.8 Security issues 3.8 Security issues
At this stage in the formation of this draft no specific security This document does not introduce any new security issues per se. A
issues have been raised. The authors welcome comments on this, detailed analysis of DHCP and DHCPv6 security is out of scope for
should such issues exist. this document.
While there is a specification for authentication for DHCP messages While there is a specification for authentication for DHCP messages
[3], the standard seems to have very few, if any, implementations. [3], the standard seems to have very few, if any, implementations.
Thus DHCP and DHCPv6 servers are still liable to be spoofed. Adding Thus DHCP and DHCPv6 servers are still liable to be spoofed. Adding
an additional protocol may give an extra avenue for attack, should an an additional protocol may give an extra avenue for attack, should an
attacker perhaps spoof a DHCPv6 server but not a DHCP server. attacker perhaps spoof a DHCPv6 server but not a DHCP server.
4. Potential solutions 4. Potential solutions
While this document did not originally intend to have solutions in Here we discuss the two broad solution strategies proposed within the
its scope, we discuss potential solution spaces in brief here in IETF dhc WG. The first is to run separate DHCP and DHCPv6 servers
order to provoke some discussion of the issues. If separate solution (with the client merging information received from both where
document(s) emerge, these notes may be removed from this document; necessary, or perhaps choosing to query a particular version first),
alternatively this document could be expanded to become a best the second is to run only a DHCPv6 server, and relay IPv4
practice guide. Comments on this are welcomed. configuration information within (new) IPv4 configuration options.
4.1 Separate DHCP servers 4.1 Separate DHCP servers
One solution is to run separate DHCP and DHCPv6 servers. These may One solution is to run separate DHCP and DHCPv6 servers. These may
or may not be run on the same physical node. The information served or may not be run on the same physical node. The information served
from the DHCP servers could be generated from a single database from the DHCP servers could be generated from a single database
instance for consistency. instance for consistency.
In this approach, some best practice guidance is required for how In this approach, some best practice guidance is required for how
multiple responses are handled or merged. Administrators have the multiple responses are handled or merged. Administrators have the
onus to maintain consistency (e.g. scripts may generate common DHCP onus to maintain consistency (for example, scripts may generate
and DHCPv6 configuration files). common DHCP and DHCPv6 configuration files).
In some cases, inconsistencies may not matter. In a simple case, an In some cases, inconsistencies may not matter. In a simple case, an
NTP server will give the same time whether accessed by IPv4 or IPv6. NTP server will give the same time whether accessed by IPv4 or IPv6.
Even if different recursive DNS servers are offered via DHCP or Even if different recursive DNS servers are offered via DHCP or
DHCPv6, those name servers will provide the same response to a given DHCPv6, those name servers will provide the same response to a given
query. The order of DNS servers in a node's configuration is not query. The order of DNS servers in a node's configuration is not
important, unless DNS load balancing is required. important, unless DNS load balancing is required.
In other cases, inconsistencies may be an issue, e.g. where lists of In other cases, inconsistencies may be an issue, for example where
values are returned, an algorithm is needed for list merger (e.g. lists of values are returned, an algorithm is needed for list merger
"alternate, DHCPv6 first"). Or there may be incompatible (e.g. "alternate, DHCPv6 first"). Or there may be incompatible
configuration values where, for example, DHCPv6 supplies domain names configuration values where, for example, DHCPv6 supplies domain names
(such the SMTP or POP servers) whereas DHCPv4 provided only IPv4 (such the SMTP or POP servers) whereas DHCPv4 provided only IPv4
addresses. addresses.
In the case of separate servers, there are some options like DNS In the case of separate servers, there are some options like DNS
search path, that aren't used in a specific IP protocol context. search path, that aren't used in a specific IP protocol context.
The multiple server approach will have some simplifications. The The multiple server approach will have some simplifications. The
DHCPv4 and DHCPv6 servers may provide the same value for a particular DHCPv4 and DHCPv6 servers may provide the same value for a particular
parameter, in which case there is no conflict. In some cases the parameter, in which case there is no conflict. In some cases the
value may be different, but the effect should be the same (e.g. NTP value may be different, but the effect should be the same (such as an
server). The crux of the issue is to identify where differences may NTP server). The crux of the issue is to identify where differences
occur and where these differences will have an impact on node may occur and where these differences will have an impact on node
behaviour. behaviour.
One possible solution is to have per-host preferences, or an ordered One possible solution is to have per-host preferences, or an ordered
list of preferences, e.g. "use manually configured"", "prefer list of preferences, for example "use manually configured"", "prefer
DHCPv4", or "prefer DHCPv6"", assuming the host can act based upon DHCPv4", or "prefer DHCPv6"", assuming the host can act based upon
which protocol is used. It is then up to the site administrator to which protocol is used. It is then up to the site administrator to
ensure values returned from either DHCP are consistent (a principle ensure values returned from either DHCP are consistent (a principle
which extends if other methods are used, e.g. NIS or SLP). which extends if other methods are used, such as NIS or SLP).
4.2 Single DHCPv6 server 4.2 Single DHCPv6 server
There is an argument for not having to configure and operate both There is an argument for not having to configure and operate both
DHCP and DHCPv6 servers in a dual-stack site environment. The use of DHCP and DHCPv6 servers in a dual-stack site environment. The use of
both servers may also lead to some redundancy in the information both servers may also lead to some redundancy in the information
served. Thus one solution may be to modify DHCPv6 to be able to served. Thus one solution may be to modify DHCPv6 to be able to
return IPv4 information. This solution is hinted at in the DHCPv6 return IPv4 information. This solution is hinted at in the DHCPv6
[4] specification: "If there is sufficient interest and demand, [4] specification: "If there is sufficient interest and demand,
integration can be specified in a document that extends DHCPv6 to integration can be specified in a document that extends DHCPv6 to
skipping to change at page 8, line 9 skipping to change at page 8, line 17
get configured with it, and that simplicity and predicatability comes get configured with it, and that simplicity and predicatability comes
from using a single server over a single transport. IPv4-capable from using a single server over a single transport. IPv4-capable
hosts will likely remain for at least 10 years, probably much longer; hosts will likely remain for at least 10 years, probably much longer;
do we want dual-stack hosts (which will become the norm) to do both do we want dual-stack hosts (which will become the norm) to do both
DHCPv4 and DHCPv6 forever while dual-stack? If you need both servers DHCPv4 and DHCPv6 forever while dual-stack? If you need both servers
to configure interfaces with addresses, and get other configuration, to configure interfaces with addresses, and get other configuration,
then you rely on two separate protocols to work (servers and relays, then you rely on two separate protocols to work (servers and relays,
etc) in order for the host to behave correctly. etc) in order for the host to behave correctly.
This approach may require the listing of a mix of IPv4 and IPv6 This approach may require the listing of a mix of IPv4 and IPv6
addresses for an option. This should be considered when new IPv6 addresses for an option. This could then be considered when new IPv6
options are introduced. There could be just two options needed, one options are introduced. There could be just two options needed, one
new option for the address delegation, and one for doing new option for the address delegation, and one for doing
encapsulation. encapsulation.
Also, there are a number of paradigms in DHCPv6 that we miss in Also, there are a number of paradigms in DHCPv6 that we miss in
DHCPv4, e.g. going away from using MAC addresses for per-host DHCPv4, such as going away from using MAC addresses for per-host
address assignment but instead using DUIDs/IAIDs, etc (although there address assignment but instead using DUIDs/IAIDs, etc (although there
is ongoing work to provide DUIDs for DHCPv4 [6]). is ongoing work to provide DUIDs for DHCPv4 [6]).
However, there are a number of potential problems with this approach: However, there are a number of potential problems with this approach:
o IPv4-only nodes would not have any DHCP service available to them; o IPv4-only nodes would not have any DHCP service available to them;
such an approach is only possible in a fully dual-stack such an approach is only possible in a fully dual-stack
environment. environment.
o The client node may then be IPv6-only and receiving IPv4 o The client node may then be IPv6-only and receiving IPv4
skipping to change at page 8, line 41 skipping to change at page 8, line 49
o What happens if there are DHCPv6 servers that don't return IPv4 o What happens if there are DHCPv6 servers that don't return IPv4
information? Does this mean the client can't run IPv4 (since it information? Does this mean the client can't run IPv4 (since it
won't do DHCPv4)? won't do DHCPv4)?
o If IPv4 information is served from a DHCPv6 server as well as an o If IPv4 information is served from a DHCPv6 server as well as an
IPv4 DHCP server, IPv4 address space will need to be allocated to IPv4 DHCP server, IPv4 address space will need to be allocated to
both servers, fragmenting the potentially precious IPv4 global both servers, fragmenting the potentially precious IPv4 global
address resource for the site. address resource for the site.
4.3 Administrative and other areas 4.3 Optimising for failure with lists of addresses
There is a generic issue with any option that includes a list of
addresses of servers (such as DNS server addresses). The list is
offered to cater for resilience, such as whether the listed server
itself fails, or connectivity to the server fails. If the client
does not know the cause of failure, its optimal strategy is to try a
different server, via a different protocol. The problem today is
that the IPv4 list is returned via DHCPv4 and the IPv6 list via
DHCPv6, and the client has no way to really "try a different server",
since that information is lost by the protocol, even though it may be
known by the server.
Just putting merging heuristics in the client cannot provide the best
behaviour, since information is lost. By comparison, if IPv4-mapped
addresses were included in the DHCPv6 option along with IPv6
addresses, the DHCP server can give an intelligent order that takes
into account which addresses are of the same DNS/whatever server.
IPv6-only clients have to know to discard the IPv4-mapped addresses
in this solution, and it's much easier to solve this in the
combined-DHCP-server case than in the two-server case.
One can argue this is only an optimisation, and in many cases the
list has only two elements, so the "next" choice is forced. However,
this particular issue highlights the subtleties of merging responses
from separate servers.
4.4 Administrative and other areas
There are also administrative issues or best practice that could be There are also administrative issues or best practice that could be
promoted. For example, it may be recommended that sites do not split promoted. For example, it may be recommended that sites do not split
their DNS name space for IPv6-specific testbeds. their DNS name space for IPv6-specific testbeds.
It may be worth considering whether separate manual configuration It may be worth considering whether separate manual configuration
files should be kept for IPv4 and IPv6 settings, e.g. separate /etc/ files should be kept for IPv4 and IPv6 settings, such as separate /
resolv.conf files for DNS settings on Unix systems. However, this etc/resolv.conf files for DNS settings on UNIX systems. However,
seems a complex solution that should be better solved by other more this seems a complex solution that should be better solved by other
generalised methods. more generalised methods.
Some differences in DHCP and DHCPv6 may not be reconciled, but may Some differences in DHCP and DHCPv6 may not be reconciled, but may
not need to be, e.g. different ways to assign addresses by DUID in not need to be, such as different ways to assign addresses by DUID in
DHCPv6, or the non-aligned option numbers for DHCP and DHCPv6. DHCPv6, or the non-aligned option numbers for DHCP and DHCPv6.
5. Summary 5. Summary
There are a number of issues in the operation of DHCP and DHCPv6 There are a number of issues in the operation of DHCP and DHCPv6
servers for nodes in dual-stack environments that should be servers for nodes in dual-stack environments that should be
clarified. While some differences in the protocols may not be clarified. While some differences in the protocols may not be
reconciled, there may not be a need to do so. However, for general reconciled, there may not be a need to do so. However, for general
operation some best practice should be agreed, the principle choice operation some best practice should be agreed, the principle choice
being whether separate DHCP and DHCPv6 servers should be maintained being whether separate DHCP and DHCPv6 servers should be maintained
by a site, or whether DHCPv6 should be extended to carry IPv4 by a site, or whether DHCPv6 should be extended to carry IPv4
configuration settings for dual-stack nodes. configuration settings for dual-stack nodes.
It can certainly be argued that until a site is completely It can certainly be argued that until a site is completely
dual-stack, an IPv4 DHCP service will always be required (e.g. while dual-stack, an IPv4 DHCP service will always be required (for example
there are still legacy printers, IP webcams or devices which still while there are still legacy printers, IP webcams or devices which
configure via DHCPv4), and a single IPv6 transport DHCP server still configure via DHCPv4), and a single IPv6 transport DHCP server
offering configuration information for both protocols will then not offering configuration information for both protocols will then not
be sufficient. In that case, there is a good rationale for focusing be sufficient. In that case, IPv4 DHCP is required, and thus there
effort on how to combine the information received from separate IPv4 is a good rationale for focusing effort on how to combine the
DHCP and (stateless) DHCPv6 servers. information received from separate IPv4 DHCP and (stateless) DHCPv6
servers.
In theory, it should be relatively straightforward to write a In theory, it should be relatively straightforward to write a
configuration manager that would accept a single configuration configuration manager that would accept a single configuration
specification from the service manager and distribute the correct specification from the service manager and distribute the correct
(and consistent) configurations to the DHCPv4 and DHCPv6 servers (and consistent) configurations to the DHCPv4 and DHCPv6 servers
(whether on the same host or not). In this case, maintaining (whether on the same host or not). In this case, maintaining
coordinated configurations in two servers is an interface issue, not coordinated configurations in two servers is an interface issue, not
a protocol issue. The question then is whether the client has all a protocol issue. The question then is whether the client has all
the information it needs to make reasonable choices. We are aware of the information it needs to make reasonable choices. We are aware of
one implementation of separate DHCPv4 and DHCPv6 clients that is one implementation of separate DHCPv4 and DHCPv6 clients that is
using a preference option for assisting client-side merging of the using a preference option for assisting client-side merging of the
received information. received information.
Another issue for discussion is whether a combined DHCP service only Another issue for discussion is whether a combined DHCP service only
available over IPv6 transport is a desirable longer-term goal for available over IPv6 transport is a desirable longer-term goal for
networks containing only dual-stack or IPv6-only nodes (or IPv4-only networks containing only dual-stack or IPv6-only nodes (or IPv4-only
nodes where DHCPv4 is not needed). The transition to the long-term nodes where DHCPv4 is not needed). The transition to the long-term
position may easily take more than 10 years. position may easily take more than 10 years.
This work has overlap with multihoming and multi-interface On reflection on the above observations, it was the strong concensus
configuration issues. It is also interwoven with the Detecting of the dhc WG to adopt the two-server approach (separate DHCP and
Network Attachment area, e.g. where a node may move from an DHCPv6 servers) in favour to a combined, single server returning IPv4
IPv4-only network to a dual-stack network, or vice versa. Both information over IPv6. The two servers may be co-located on a single
aspects may be best abstracted for discussion in the IETF multi6 and node, and may have consistent configuration information generated
dna WGs for discussion. from a single asset database.
The authors also noted that the original working title of the draft Having reached this concensus, future work is now required to
was not as appropriate as it might be; we have thus renamed it "DHCP: determine best practice for merging information from multiple
IPv4 and IPv6 Dual-Stack Issues". We are open to further renaming if servers, including merger of lists of addresses where options carry
comments warrant it. such information.
There is not a full consensus in the DHC WG on solutions for the DHCP As a footnote, we note that this work has overlap with multihoming
dual-stack configuration issue at present. This text is intended to and multi-interface configuration issues. It is also interwoven with
provoke discussion towards a consensus, and it may then document that the Detecting Network Attachment area, for example where a node may
consensus and the reasons behind it for future reference. move from an IPv4-only network to a dual-stack network, or vice
versa. Both aspects may be best abstracted for discussion and
progression in the respective IETF multi6 and dna WGs, in parallel
with the two-server progression in the dhc WG.
6. Security Considerations 6. Security Considerations
There are no security considerations in this problem statement per There are no security considerations in this problem statement per
se, as it does not propose a new protocol. se, as it does not propose a new protocol.
7. Acknowledgements 7. Acknowledgements
The authors thank the following people for input to this draft: The authors thank the following people for input to this document:
Bernie Volz, AK Vijayabhaskar, Ted Lemon, Ralph Droms, Robert Elz, Bernie Volz, AK Vijayabhaskar, Ted Lemon, Ralph Droms, Robert Elz,
Changming Liu, Margaret Wasserman and Greg Daley. The draft may not Changming Liu, Margaret Wasserman, Dave Thaler and Greg Daley. The
fully reflect the views of each of these individuals. document may not necessarily fully reflect the views of each of these
individuals.
The authors would also like to thank colleagues on the 6NET project The authors would also like to thank colleagues on the 6NET project
for contributions to this draft. for contributions to this document.
8 Informative References 8 Informative References
[1] Droms, R., "Dynamic Host Configuration Protocol", RFC 2131, [1] Droms, R., "Dynamic Host Configuration Protocol", RFC 2131,
March 1997. March 1997.
[2] Thomson, S. and T. Narten, "IPv6 Stateless Address [2] Thomson, S. and T. Narten, "IPv6 Stateless Address
Autoconfiguration", RFC 2462, December 1998. Autoconfiguration", RFC 2462, December 1998.
[3] Droms, R. and W. Arbaugh, "Authentication for DHCP Messages", [3] Droms, R. and W. Arbaugh, "Authentication for DHCP Messages",
RFC 3118, June 2001. RFC 3118, June 2001.
[4] Droms, R., Bound, J., Volz, B., Lemon, T., Perkins, C. and M. [4] Droms, R., Bound, J., Volz, B., Lemon, T., Perkins, C. and M.
Carney, "Dynamic Host Configuration Protocol for IPv6 (DHCPv6)", Carney, "Dynamic Host Configuration Protocol for IPv6 (DHCPv6)",
RFC 3315, July 2003. RFC 3315, July 2003.
[5] Droms, R., "Stateless Dynamic Host Configuration Protocol (DHCP) [5] Droms, R., "Stateless Dynamic Host Configuration Protocol (DHCP)
Service for IPv6", RFC 3736, April 2004. Service for IPv6", RFC 3736, April 2004.
[6] Lemon, T., "Node-Specific Client Identifiers for DHCPv4", [6] Lemon, T., "Node-Specific Client Identifiers for DHCPv4",
draft-ietf-dhc-3315id-for-v4-02 (work in progress), February draft-ietf-dhc-3315id-for-v4-03 (work in progress), July 2004.
2004.
Authors' Addresses Authors' Addresses
Tim Chown Tim Chown
University of Southampton University of Southampton
School of Electronics and Computer Science School of Electronics and Computer Science
Southampton, Hampshire SO17 1BJ Southampton, Hampshire SO17 1BJ
United Kingdom United Kingdom
EMail: tjc@ecs.soton.ac.uk EMail: tjc@ecs.soton.ac.uk
 End of changes. 

This html diff was produced by rfcdiff 1.23, available from http://www.levkowetz.com/ietf/tools/rfcdiff/