One document matched: draft-ietf-dhc-option-guidelines-07.xml


<?xml version='1.0' ?>

<!-- $Id: draft-ietf-dhc-option-guidelines.xml,v 1.17 2010/03/06 01:03:43 dhankins Exp $ -->

<?rfc toc="yes"?>
<?rfc sortrefs="yes"?>
<?rfc symrefs="yes"?>

<?rfc compact="yes"?> <?rfc subcompact="no"?>

<!DOCTYPE rfc SYSTEM 'rfc2629bis.dtd' [
  <!ENTITY rfc1035 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.1035.xml'>
  <!ENTITY rfc2131 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.2131.xml'>
  <!ENTITY rfc2132 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.2132.xml'>
  <!ENTITY rfc2241 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.2241.xml'>
  <!ENTITY rfc2242 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.2242.xml'>
  <!ENTITY rfc3011 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.3011.xml'>
  <!ENTITY rfc3046 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.3046.xml'>
  <!ENTITY rfc3118 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.3118.xml'>
  <!ENTITY rfc3315 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.3315.xml'>
  <!ENTITY rfc3319 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.3319.xml'>
  <!ENTITY rfc3397 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.3397.xml'>
  <!ENTITY rfc3442 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.3442.xml'>
  <!ENTITY rfc3495 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.3495.xml'>
  <!ENTITY rfc3527 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.3527.xml'>
  <!ENTITY rfc3634 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.3634.xml'>
  <!ENTITY rfc3646 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.3646.xml'>
  <!ENTITY rfc3898 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.3898.xml'>
  <!ENTITY rfc3925 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.3925.xml'>
  <!ENTITY rfc3942 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.3942.xml'>
  <!ENTITY rfc4030 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.4030.xml'>
  <!ENTITY rfc4075 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.4075.xml'>
  <!ENTITY rfc4174 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.4174.xml'>
  <!ENTITY rfc4280 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.4280.xml'>
  <!ENTITY rfc4702 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.4702.xml'>
  <!ENTITY rfc4704 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.4704.xml'>
  <!ENTITY rfc4833 PUBLIC ''
	'http://xml.resource.org/public/rfc/bibxml/reference.RFC.4833.xml'>
]>
<rfc ipr="trust200902" category="info"
     docName="draft-ietf-dhc-option-guidelines-07">
  <front>
    <title abbrev="DHCP Guidelines">Guidelines for Creating New DHCP
	Options</title>

    <author initials="D." surname="Hankins" fullname="David W. Hankins">
	<organization abbrev="Google">Google, Inc.</organization>

	<address>
	  <postal>
	    <street>1600 Amphitheatre Parkway</street>
	    <city>Mountain View</city>
	    <code>94043</code>
	    <region>CA</region>
	    <country>USA</country>
	  </postal>

	  <email>dhankins@google.com</email>
	</address>
    </author>

    <date/>

    <area>Internet</area>
    <workgroup>Dynamic Host Configuration Working Group</workgroup>

    <keyword>DHCP</keyword>
    <keyword>option guidelines</keyword>
    <keyword>option guidance</keyword>
    <keyword>option format</keyword>
    <keyword>atomic dhcp</keyword>

    <abstract>
	<t>This document seeks to provide guidance to prospective DHCP
	Option authors, to help them in producing option formats that
	are easily adoptable.</t>
    </abstract>
  </front>

  <middle>
    <section title="Introduction">
      <t>Most protocol developers ask themselves if a protocol will work,
	or work efficiently.  These are important questions, but another
	less frequently considered question is whether the proposed protocol
	presents itself needless barriers to adoption by deployed
	software.</t>

      <t><xref target="RFC2131">DHCPv4</xref> and
	<xref target="RFC3315">DHCPv6</xref> software implementors are not
	merely faced with the task of a given option's format on the wire.
	The option must "fit" into every stage of the system's process, from
	the user interface where configuration is entered to the machine
	interfaces where configuration is consumed.  To help understand the
	potential implementation challenges of any new DHCP Option, <xref
	target="isc">one implementation's approach to tackling DHCP
	Option formats</xref> has been included as an Appendix.</t>

      <t>Another more frequently overlooked aspect of rapid adoption
	is the question:  Would the option would require operators to be
	intimately familiar with the option's internal format in order to
	make use of it?  Most DHCP software provides a facility for "unknown
	options" at the time of publication to be configured by hand by an
	operator.  But if doing so requires extensive reading (more than can
	be covered in a simple FAQ for example), it inhibits adoption.</t>

      <t>So although a given solution would work, and might even be space,
	time, or aesthetically optimal, a given option is presented with a
	series of ever-worsening challenges to be adopted;

	<list style="symbols">
	  <t>If it doesn't fit neatly into existing config files.</t>
	  <t>If it requries new source code changes to be adopted, and
	     hence upgrades of deployed software.</t>
	  <t>If it does not share its deployment fate in a general manner with
	     other options, standing alone in requiring code changes or
	     reworking configuration file syntaxes.</t>
	</list>
      </t>

      <t>There are many things DHCP option authors can do to form
	DHCP Options to stay off this list entirely, or failing that, to make
	software implementors lives easier and improve its chances for
	widespread adoption.</t>
    </section>

    <section title="When to Use DHCP">
      <t>Principally, DHCP carries configuration parameters for its clients.
	Any knob, dial, slider, or checkbox on the client system, such as
	"my domain name servers", "my hostname", or even "my shutdown
	temperature" are candidates for being configured by DHCP.</t>

      <t>The presence of such a knob isn't enough, because DHCP also presents
	the extension of an administrative domain - the operator of the network
	to which the client is currently attached.  Someone runs not only
	the local switching network infrastructure that the client is
	directly (or wirelessly) attached to, but the various methods of
	accessing the external Internet via local assist services that
	network must also provide (such as domain name servers, or routers).
	This means that in addition to the existence of a configuration
	parameter, one must also ask themselves if it is reasonable for
	this parameter to be set by the directly attached network's
	administrators.</t>

      <t>Bear in mind that the client still reserves the right to ignore
	values received via DHCP (for example, due to having a value
	manually configured by its own operator), and that at least one main
	use case for DHCP is the corporate enterprise.  So even if the local
	Net Cafe's operator is not a likely source of the candidate
	configuration, there may be other DHCP servers in a client's
	lifetime which are.</t>
    </section>

    <section title="General Principles">
      <t>The primary principle to follow in order to enhance an option's
	adoptability is certainly simplification.  But more specifically,
	to create the option in such a way that it should not require
	any new or special case software to support.  If old software
	currently deployed and in the field can adopt the option
	through supplied configuration conveniences then it's fairly
	well assured that new software can easily formally adopt it.</t>

      <t>There are at least two classes of DHCP options:  A bulk class of
	options which are provided explicitly to carry data from one side
	of the DHCP exchange to the other (such as nameservers, domain
	names, or time servers), and a protocol class of options which
	require special processing on the part of the DHCP software or are
	used during special processing (such as the FQDN options
	(<xref target="RFC4702"/>, <xref target="RFC4704"/>),
	<xref target="RFC2132">DHCPv4 message type option</xref>, link
	selection options (<xref target="RFC3011"/>, <xref target="RFC3527"/>),
	and so forth; these options carry data that is the result of a routine
	in some DHCP software).</t>

      <t>The guidelines laid out here should be understood to be relaxed
	for the protocol class of options.  Wherever special-case-code is
	already required to adopt the DHCP option, it is substantially
	more reasonable to format the option in a less generic fashion,
	if there are measurable benefits to doing so.</t>
    </section>

    <section title="Reusing Other Options">
      <t>In DHCPv4, there are now nearly one hundred and thirty options,
	at least as IETF standards, which might be used as an example.
	There is also one <xref target="RFC2132">handy document</xref>
	containing many option definitions.</t>

      <t>There is a tradeoff between the adoptability of previously defined
	option formats, and the advantages new or specialized formats can
	provide.  In the balance, it is usually preferrable to reuse 
	previously used option formats.</t>

      <t>However, it isn't very practical to consider the bulk of DHCP options
	already allocated, and consider which of those solve a similar
	problem.  So, the following list of common option format fragments is
	provided as a shorthand.  Please note that it is not complete in terms
	of exampling every option format ever devised...it is only a list of
	option format fragments which are used in two or more options.</t>

      <texttable anchor='fragments' title='Common Option Fragments'>

	<!-- Note the Size column is right-alingned to provide decimal
	  -- alignment (4/16, 4 and 6 are on the same text column). -->
	<ttcol align='right'>Fragment</ttcol>
	<ttcol align='right'>Size</ttcol>
	<ttcol align='left'>Types of Uses</ttcol>

	<c>ipv4-address</c>
	<c>4</c>
	<c><xref target="RFC2132">Default gateway, requested address,
	   subnet mask</xref>, addresses of servers (<xref target="RFC2132"/>,
	   <xref target="RFC2241"/>, <xref target="RFC2242"/>,
	   <xref target="RFC3495"/>, <xref target="RFC3634"/>,
	   <xref target="RFC4174"/>), as a component in a
	   <xref target="RFC3442">list of routes</xref>.</c>

	<c>ipv6-address</c>
	<c>16</c>
	<c><xref target="RFC3315">DHCPv6 server unicast address</xref>,
	   addresses of servers (<xref target="RFC3319"/>,
	   <xref target="RFC3646"/>, <xref target="RFC3898"/>,
	   <xref target="RFC4075"/>, <xref target="RFC4280"/>).</c>

	<c>32-bit integer</c>
	<c>4</c>
	<c>Signed or unsigned varieties.
	   Used for <xref target="RFC2132">timezone time offset</xref>
	   (deprecated by <xref target="RFC4833"/>).  Other uses
	   for host configuration values such as <xref target="RFC2132">path
	   MTU aging timeouts, ARP cache timeouts, TCP keepalive
	   intervals</xref>.  Also used by the DHCPv4 protocol for relative
	   times, and times since epoch.</c>

	<c>16-bit integer</c>
	<c>2</c>
	<c>Client configuration parameters, such as <xref target="RFC2132">MTU,
	   maximum datagram reassembly limits, the DHCPv4 maximum message
	   size</xref>, or the <xref target="RFC3315">elapsed time
	   option</xref> in DHCPv6.</c>

	<c>8-bit integer</c>
	<c>1</c>
	<c>Used for host configuration parameters, such as the
	   <xref target="RFC2132">default IP TTL, default TCP TTL, NetBIOS
	   node type</xref>.  Also used for protocol features, such as the
	   DHCPv4 Option Overload (as flags), DHCP Message Type (as an
	  enumeration) or <xref target="RFC3315">DHCPv6 Preference</xref>.</c>

	<c>NVT-ASCII Text</c>
	<c>unlim</c>
	<c>This is the kitchen sink of common fragments.  Common uses are for
	   filenames (such as TFTP paths), host or domain names (but this
	   should be discouraged), or protocol features such as textual
	   messages such as verbose error indicators.  Since the size of
	   this format cannot be determined (it is not NULL terminated),
	   it consumes any remaining space in the option.</c>

	<c><xref target="RFC1035">DNS Wire Format Domain Name List</xref></c>
	<c>unlim</c>
	<c>Presently used for 'domain search' lists in both DHCPv4
	   <xref target="RFC3397"/> and DHCPv6 <xref target="RFC3646"/>, but
	   also used in DHCPv6 for any host or domain name.  A field
	   formatted this way may have a determinate length if the number of
	   root labels is limited, but use of this format as being a
	   determinate length should be discouraged in DHCPv4, less so in
	   DHCPv6.</c>

	<c>'suboption' encapsulation</c>
	<c>unlim</c>
	<c>The <xref target="RFC3046">Relay Agent Information Option</xref>,
	   <xref target="RFC2132">vendor options</xref>, Vendor
	   Identified Vendor SubOptions (<xref target="RFC3925"/>,
	   <xref target="RFC3315"/>).  Commonly used for situations
	   where the full format cannot be known initially, such as
	   where there seems to be some room for later protocol work to
	   expand the amount of information carried, or where the full
	   extent of data carried is defined in a private specification
	   (such as with vendor options).  Encapsulations do not use
	   'PAD' and 'END' options in DHCPv4, and there are no such options
	   in DHCPv6, so this format also is of indeterminate length.</c>
      </texttable>

      <t>The easiest approach to manufacturing trivially deployable DHCP
	Options is to assemble the option out of whatever common fragments
	fit - possibly allowing a group of fragments to repeat to fill
	the remaining space (if present) and so provide multiple values.
	Place all fixed size values at the start of the option, and any
	variable/indeterminate sized value at the tail end of the option.</t>

      <t>This estimates that implementations will be able to reuse code paths
	designed to support the other options.</t>
    </section>

    <section title="Avoid Conditional Formatting">
      <t>Placing a octet at the start of the option which informs the
	software how to process the remaining octets of the option may
	appear simple to the casual observer.  But the only conditional
	formatting methods that are in widespread use today are 'protocol'
	class options.  So conditional formatting requires new code to be
	written, as well as introduces an implementation problem; as it
	requires that all speakers implement all current and future conditional
	formats.</t>

      <t>Conditional formatting is absolutely not recommended, except in cases
	where the DHCP option has already been deployed experimentally, and
        all but one conditional format is deprecated.</t>
    </section>

    <section title="Avoid Aliasing">
      <t>Options are said to be aliases of each other if they provide input to
	the same configuration parameter.  A commonly proposed example is to
	configure the location of some new service ("my foo server") using a
	binary IP address, a domain name field, and a URL.  This kind of
	aliasing is undesirable, and is best avoided.</t>

      <t>In this case, where three different formats are supposed, it triples
	the work of the software involved, requiring support for not merely
	one format, but support to produce and digest all three.  Since
	clients cannot predict what values the server will provide, they
	must request all formats...so in the case where the server is
	configured with all formats, DHCP option space is wasted on option
	contents that are redundant.</t>

      <t>It also becomes unclear which types of values are mandatory, and how
	configuring some of the options may influence the others.  For example,
	if an operator configures the URL only, should the server synthesize
	a domain name and IP address?</t>

      <t>A single configuration value on a host is probably presented to the
	operator (or other software on the machine) in a single field or
	channel.  If that channel has a natural format, then any alternative
	formats merely make more work for intervening software in providing
	conversions.</t>

      <t>So the best advice is to choose the one method that best fulfills
	the requirements, be that for simplicity (such as with an IP address
	and port pair), late binding (such as with DNS), or completeness
	(such as with a URL).</t>

      <t>On the specific subject of desiring to configure a value using a
	Fully Qualified Domain Name instead of a binary IP address, note that
	most DHCP server implementations will happily accept a Domain Name
	entered by the administrator, and use DNS resolution to render binary
	IP addresses in DHCP replies to clients.  Consequently, consider the
	extra packet overhead incurred on the client's end to perform DNS
	resolution itself.  The client may be operating on a battery
	and packet transmission is a non-trivial use of power, and the extra
	RTT delays the client must endure before the service is configured are
	at least two factors to consider in making a decision on format.</t>
    </section>

    <section title="Considerations for Creating New Formats">
      <t>If the option simply will not fit into any existing work by using
	fragments, the last recourse is to create a new format to fit.</t>

      <t>When doing so, it is not enough to gauge whether or not the option
	format will work in the context of the option presently being
	considered.  It is equally important to consider if the new format's
	fragments might reasonably have any other uses, and if so, to create
	the option with the foreknowledge that its parts may later become a
	common fragment.</t>

      <t>One specific consideration to evaluate is whether or not options
	of a similar format would need to have multiple or single values
	encoded (whatever differs from the current option), and how that
	might be accomplished in a similar format.</t>
    </section>

    <section title="The Dangers of Sub Options">
      <t>Some DHCP options, such as the <xref target="RFC3046">DHCPv4 Relay
	Agent Information Option</xref> are defined to contain a series of
	DHCP options, possibly using code tags specific to that option (but
	not in some limited "protocol feature" cases in
	<xref target="RFC3315">DHCPv6</xref>).  These are commonly referred to
	as Encapsulated Option Spaces or more simply, Sub Options.</t>

      <t>Sub options seem very attractive, because they allow the encoding of
	multiple variable length fields within the single "parent" option.
	However, DHCP software will only include these options on an "all or
	nothing" basis, there is no well deployed mechanism for "Sub Option
	Parameter Request Lists" (although some defined sub-option spaces,
	such as for DOCSIS, do describe sub-option scoped PRL analogues), and
	the software will not include the entire option if there is not
	sufficient space.</t>

      <t>Consequently, it is not advisable to group options that may not
	be requested at the same time by the same client under an encapsulated
	space.</t>

      <t>Another attraction sub options present is ease of extending the
	configuration value for later, related configuration.  This must be
	weighed against the cost associated with asking IANA to maintain the
	space's internally assigned option codes.  Generally, the cost to IANA
	is greater, as it is unlikely that options will be later extended.</t>

      <t>The use of sub-options is not a solution to aliasing problems.
	Sub-options that contain multiple configuration values that alias the
	same configuration element actually makes matters worse.  The only
	solution to aliasing problems is to select a single option format, or
	where that is literally impossible, to use multiple DHCP options.  In
	this way, clients may place only the options they support on their
	parameter request list, in the order they support them.  Later protocol
	maintenance may incorporate a means to select a single DHCP option code
	out of a list of aliased options, so do not concern yourself with
	packet space issues arising from receiving all the aliases.</t>

      <t>Additionally, DHCPv4 <xref target="fragmentation">option concatenation
	</xref> has not been defined in any DHCPv4 sub-options space.
	Currently there is some DHCP software which does concatenate multiple
	DHCP options present in a sub-option space.  There is also software
	that treats multiple DHCP option codes present in a sub-option as
	individual single options.  So there is no reliably predictable default
	behaviour.</t>

      <t>Because no sub-options space has yet been defined that includes the
	possibility of having more than one instance of an option of the same
	code, any attempt to do so is discouraged.</t>
    </section>

    <section anchor="fragmentation" title="Option Size">
      <t><xref target="RFC2131">DHCPv4</xref> options payload space is limited,
	as there are a number of unaddressed deployment problems with DHCPv4
	packet sizes.  The end result is that you should build your option to
	the assumption that the packet will be no larger than 576 octets.  This
	means that the options payload space will be 312 octets, which you will
	have to share with other options.  This space can be extended by
	making use of <xref target="RFC2132">Option Overloading</xref>,
	which allows the use of the BOOTP FILE and SNAME header fields for
	carrying DHCPv4 options (adding 192 octets), but these header fields
	will not be available for overloading if they have been configured
	to carry a value.</t>

      <t><xref target="RFC3315">DHCPv6</xref> is much better off.  First,
	through its use of link-local addresses, it steps aside many of the
	deployment problems that plague DHCPv4, and looks a great deal more
	like any other UDP based application; oblivious to packet sizes up
	to 64KB.  Second, RFC 3315 explicitly refers readers to RFC 2460 Section
	5, which describes an MTU of 1280 octets and a minimum fragment
	reassembly of 1500 octets.  It's much more feasible to suggest that
	DHCPv6 is capable of having larger options deployed over it, and at
	least no common upper limit is yet known to have been encoded by its
	implementors.  It is impossible to describe any fixed limit that
	cleanly divides those too big from the workable.</t>

      <t>So in either protocol, it is advantageous to prefer option formats
	which contain the desired information in the smallest form factor that
	solves the requirements.  One example is to use a 4-octet IPv4 address
	rather than a fully qualified domain name, because many DHCP servers
	will perform DNS resolution on configured FQDN's (so the DNS recursive
	lookup is performed anyway).  There may be motivations to use the
	fully qualified domain name anyway, such as if the intended RRSET is
	not an address, or if the client must refresh the name more frequently
	than common lease renewal periods.</t>

      <t>When it is not possible to compress the configuration contents either
	because of the simple size of the parameters, or because it is
	expected that very large configurations are valid, it may be
	preferable to use a second stage configuration.  Some examples of
	this are to provide TFTP server and pathnames, or a URL, which the
	client will load and process externally to the DHCP protocol.</t>

      <t>The DHCPv4 code and length tags are each a single octet.  As the
	length field describes the length of the DHCP option's contents (not
	including the code and length octets), any option whose contents'
	length exceeds 255 octets can not be contained in a single option.
	These 'long options' will simply be fragmented into multiple options
	within the packet.  DHCP software processing these fragments will
	concatenate them, in the order they appear as defined by
	<xref target="RFC2131"/>, prior to evaluating their
	contents.  This is an important distinction that is sometimes
	overlooked by authors - these multiple options are not individually
	formatted to convey one unit of information precisely as you have
	defined, but rather one option that has been split along any arbitrary
	octet boundary into multiple containers.  When documenting an example,
	then, try to make sure that the division point you select as an
	example does not lie on a clean division of your option contents -
	place it at an offset so as to reinforce that these values must be
	concatenated rather than processed individually.</t>

      <t>DHCPv4 option fragments are a basic protocol feature, so there
	usually is no reason to mention this feature in new option definitions,
	and no requirement for every option definition to be presented as a
	series of fragments.  It is only recommended to reinforce the existence
	of DHCP option fragmentation when the potential for large options is
	likely.  In this case, try to choose a large example data value.</t>

      <t>Note that option fragmentation is also a very common side-effect of
	running out of options space, and moving to overloaded FILE or SNAME
	fields.  Although the option may be considerably shorter than 255
	octets, if it does not fit in the remaining space then software may
	consume all remaining options space with one option fragment, and
	place the remainder in an overloaded field.</t>

      <t>Primarily it is important to remember that DHCPv4 differs from DHCPv6
	on this point:  DHCPv4 can only convey one option of a given option
	code at any time - additional options (or possibly sub options, which
	do not have concisely defined semantics) of the same code will be
	concatenated together and processed at once.  DHCPv6 does allow
	multiple instances of a given option, and they are treated as distinct
	values following the defined format, however this feature is generally
	preferred to be restricted to protocol class features (such as the
	IA_* series of options); it is better to define your option as an
	array if it is possible.</t>

      <t>So remember that it is out of the question to define a case for
	multiple instances of your option in DHCPv4, and it is recommended to
	clarify (with normative language) if any DHCPv6 option may appear once
	or multiple times.</t>
    </section>

    <section title="Clients Request their Options">
      <t>The <xref target="RFC2132">DHCPv4 Parameter Request List</xref>, and
	the <xref target="RFC3315">DHCPv6 Option Request Option
	(OPTION_ORO)</xref>, are both options that serve two purposes - to
	inform the server what options the client supports and is willing to
	digest, and in what order of priority the client places those option
	contents (in the event that they will not fit in the packet, later
	options are to be dropped).</t>

      <t>It doesn't make sense for some options to appear on this Parameter
	Request List, such as those formed by elements of the protocol's
	internal workings, or are formed on either end by DHCP-level software
	engaged in some exchange of information.  When in any form of doubt,
	assume that any new option must be present on the relevant option
	request list if the client desires it.</t>

      <t>It is a frequent mistake of option draft authors, then, to create
	text that implies that a server will simply provide the new option,
	and clients will digest it.  Generally, it's best to also specify
	that clients MUST place the new option code on the relevant list
	option, clients MAY include the new option in their packets to
	servers with hints as to values they desire, and servers MAY respond
	with the option contents (if they have been so configured).</t>

      <t>Under only the most dire of circumstances should a new option
	definition require special ordering of options either in the relevant
	request option, or in the order of options within the packet.
	Although the request option does imply a priority, which might be
	processed in order, a server may shuffle options around in a DHCPv4
	packet in order to make them fit, and server software may sort DHCPv6
	options into strange orders.  There is not one universal approach.</t>
    </section>

    <section title="Security Considerations">
      <t>DHCP does have an Authentication mechanism (<xref target="RFC3118"/>,
	<xref target="RFC3315"/>, <xref target="RFC4030"/>), where it is
	possible for DHCP software to discriminate between authentic
	endpoints and men in the middle.</t>

      <t>However, at this date the mechanism is poorly deployed.  It also
	does not provide end-to-end encryption.</t>

      <t>So, while creating a new option, bear in mind that DHCP packet
	contents are always transmitted in the clear, and actual production
	use of the software will probably be vulnerable at least to
	man-in-the-middle attacks from within the network, even where the
	network itself is protected from external attacks by firewalls.  In
	particular, some DHCP message exchanges are transmitted to broadcast
	or multicast addresses that are likely broadcast anyway.</t>

      <t>If an option is of a specific fixed length, it is useful to
	remind the implementer of the option data's full length.  This is
	easily done by declaring the specific value of the 'length' tag of
	the option.  This helps to gently remind implementers to validate
	option length before digesting them into likewise fixed length
	regions of memory or stack.</t>

      <t>If an option may be of variable size (such as having indeterminate
	length fields, such as domain names or text strings), it is
	advisable to explicitly remind the implementor to be aware of
	the potential for long options.  Either define a reasonable upper
	limit (and suggest validating it), or explicitly remind the
	implementor that an option may be exceptionally long (to be prepared
	to handle errors rather than truncate values).</t>

      <t>For some option contents, "insane values" may be used to breach
	security.  An IP address field might be made to carry a loopback
	address, or local broadcast address, and depending on the protocol
	this may lead to undesirable results.  A domain name field may
	be filled with contrived contents that exceed the limitations
	placed upon domain name formatting...as this value is possibly
	delivered to "internal configuration" records of the system, it
	may be trusted, rather than validated.</t>

      <t>So it behooves an option's definition to contain any validation
	measures as can reasonably be made.</t>
    </section>

    <section title="IANA Considerations">
      <t>This document has no actions for IANA.</t>
    </section>
  </middle>

  <back>
    <references title="Informative References">
      &rfc1035;
      &rfc2131;
      &rfc2132;
      &rfc2241;
      &rfc2242;
      &rfc3011;
      &rfc3046;
      &rfc3118;
      &rfc3315;
      &rfc3319;
      &rfc3442;
      &rfc3495;
      &rfc3397;
      &rfc3527;
      &rfc3634;
      &rfc3646;
      &rfc3898;
      &rfc3925;
      &rfc3942;
      &rfc4030;
      &rfc4075;
      &rfc4174;
      &rfc4280;
      &rfc4702;
      &rfc4704;
      &rfc4833;
    </references>

    <section title="Background on ISC DHCP" anchor="isc">
      <t>The ISC DHCP software package was mostly written by Ted Lemon
	in cooperation with Nominum, Inc.  Since then, it has been given
	to Internet Systems Consortium, Inc. ("ISC") where it has been
	maintained in the public interest by contributors and ISC
	employees.</t>

      <t>It includes a DHCP Server, Relay, and Client implementation,
	with a common library of DHCP protocol handling procedures.</t>

      <t>The DHCP Client may be found on some Linux distributions, and
	FreeBSD 5 and earlier.  Variations ("Forks") of older versions of
	the client may be found on several BSDs, including FreeBSD 6 and
	later.</t>

      <t>The DHCP Server implementation is known to be in wide use by
	many Unix-based servers, and comes pre-installed on most Linux
	distributions.</t>

      <t>The ISC DHCP Software Suite has to allow:

	<list style="symbols">
	  <t>Administrators to configure arbitrary DHCP Option Wire Formats
		for options that either were not published at the time the
		software released, or are of the System Administrator's
		invention (such as <xref target="RFC3942">'Site-Local'</xref>
		options), or finally were of Vendor design
		(<xref target='RFC2132'>Vendor Encapsulated Options</xref>
		or similar).</t>

	  <t>Pre-defined names and formats of options allocated by IANA and
		defined by the IETF Standards body.</t>

	  <t>Applications deriving their configuration parameters from values
		provided by these options to receive and understand their
		content.  Often, the binary format on the wire is not
		helpful or digestable by, for example, 'ifconfig' or
		'/etc/resolv.conf'.</t>
	</list>
      </t>

      <t>So, one can imagine that this would require a number of software
	functions:

	<list style="numbers">
	  <t>To read operator-written configuration value into memory.</t>

	  <t>To write the in-memory representation into protocol wire
	  format.</t>

	  <t>To read the protocol wire format into memory.</t>

	  <t>To write the in-memory format to persistent storage (to cache
	  across reboots for example).</t>

	  <t>To write the in-memory format to a format that can be consumed
	  by applications.</t>
	</list>
      </t>

      <t>If every option were formatted differently and uniquely, then we
	would have to write 5 functions for every option.  As there is
	the potential for as many as 254 DHCPv4 options, or 65536 DHCPv6
	options, not to mention the various encapsulated spaces
	("suboptions"), this is not scalable.</t>

      <t>One simple trick is to make the in-memory format the same as
	the wire format.  This reduces the number of functions required
	from 5 to 3.  This is not always workable - sometimes an
	intermediate format is required, but it is a good general case.</t>

      <t>Another simple trick is to use the same (or very nearly the same)
	format for persistent storage as is used to convey parameters to
	applications.  This reduces the number of functions again from 3
	to 2.</t>

      <t>This is still an intractable number of functions per each DHCP
	option, even without the entire DHCP option space populated.  So, we
	need a way to reduce this to small orders.</t>

      <section title="Atomic DHCP">
	<t>To accomplish these goals, a common "Format String" is used to
	describe, in abstract, all of the above.  Each octet in this format
	string represents a "DHCP Atom".  We chain these 'atoms' together,
	forming a sort of molecular structure for a particular DHCP
	option's defined format.</t>

	<t>The Configuration Syntax allows the user to construct such
	a format string without having to understand how the Atom is
	encoded on the wire, and how it is configured or presented.</t>

	<t>You can reasonably imagine that the <xref target="fragments">various
	common formats of DHCP options described above</xref> each have an
	Atom associated with it.  There are special use Atoms, such as one
	to repeat the previous Atoms indefinitely (for example, for options
	with multiple IPv4 addresses one after the other), and one which
	makes the previous Atom optional.</t>

	<t>As the software encounters a format string, it processes each
	Atom individually to read from configuration into wire format, and also
	to validate and convert wire format into output format (which with
	some small exclusions is identical to the configuration format).</t>

	<t>The format strings themselves are either hard coded by the
	software in a table of option definitions, or are compiled
	at runtime through configuration syntax generated by the
	operator.</t>

<figure><artwork>
        option <space>.<option> code <number> = <definition>;
</artwork></figure>

	<t>The <space> refers to the option space, which may be the
	DHCPv4 option space, the DHCPv6 option space, or any suboption
	space such as the DHCPv4 Relay Agent Information suboptions or
	similar.</t>

	<t>The <option> refers to the option's symbolic name within
	that space.</t>

	<t>The code <number> refers to the binary code assigned to this
	option.</t>

	<t>The <definition> is a complex statement that brings together
	DHCP Atoms, many of which are the aforementioned common formats,
	that compose this option.</t>

	<t>Below is a sample configuration for two options, whose wire
	formats are defined in <xref target="RFC2132"/>.  The Path MTU Plateau
	Table option, and the Static Routes option.</t>

<figure><artwork>
     option dhcp.path-mtu-plateau-table code 25 =
                                        array of unsigned integer 16;
     option dhcp.static-routes code 33 = array of { ip-address,
                                                    ip-address };
</artwork></figure>

	<t>Once the options' syntax configuration is out of the way, values
	to be carried in the options may be configured.  These acts are
	distinct; the previous configuration only prepares the parser system
	to accept the configuration below.  The below configuration actually
	supplies a value to be transmitted on the wire, relying on the above
	format definition.</t>

<figure><artwork>
     option dhcp.path-mtu-plataeu-table 4352, 1500, 576;
     option dhcp.static-routes 10.10.10.10 10.10.10.9,
                               10.10.10.11 10.10.10.9;
</artwork></figure>
      </section>
    </section>
  </back>
</rfc>

PAFTECH AB 2003-20262026-04-23 19:37:58