HTTP Remote Variant Selection Algorithm -- RVSA/1.0
draft-ietf-http-rvsa-v10-02

Versions: 00 01 02 rfc2296                                              
HTTP Working Group                                     Koen Holtman, TUE
Internet-Draft                              Andrew Mutz, Hewlett-Packard
Expires: August 5, 1997                                 February 5, 1997


              HTTP Remote Variant Selection Algorithm -- RVSA/1.0

                     draft-ietf-http-rvsa-v10-00.txt


STATUS OF THIS MEMO

        This document is an Internet-Draft. Internet-Drafts are
        working documents of the Internet Engineering Task Force
        (IETF), its areas, and its working groups. Note that other
        groups may also distribute working documents as
        Internet-Drafts.

        Internet-Drafts are draft documents valid for a maximum of
        six months and may be updated, replaced, or obsoleted by
        other documents at any time. It is inappropriate to use
        Internet-Drafts as reference material or to cite them other
        than as "work in progress".

        To learn the current status of any Internet-Draft, please
        check the "1id-abstracts.txt" listing contained in the
        Internet-Drafts Shadow Directories on ftp.is.co.za
        (Africa), nic.nordu.net (Europe), munnari.oz.au (Pacific
        Rim), ds.internic.net (US East Coast), or ftp.isi.edu (US
        West Coast).

        Distribution of this document is unlimited.  Please send
        comments to the HTTP working group at
        <http-wg@cuckoo.hpl.hp.com>.  Discussions of the working
        group are archived at
        <URL:http://www.ics.uci.edu/pub/ietf/http/>.  General
        discussions about HTTP and the applications which use HTTP
        should take place on the <www-talk@w3.org> mailing list.

        HTML and change bar versions of this document are available
        at <URL:http://gewis.win.tue.nl/~koen/conneg/>.


ABSTRACT

        HTTP allows web site authors to put multiple versions of the
        same information under a single URL.  Transparent content
        negotiation is a mechanism for automatically selecting the
        best version when the URL is accessed.  A remote variant
        selection algorithm can be used to speed up the transparent
        negotiation process. This document defines the remote variant
        selection algorithm with the version number 1.0.


OVERVIEW OF THE TRANSPARENT CONTENT NEGOTIATION DOCUMENT SET

   An up-to-date overview of documents related to transparent content
   negotiation is maintained on the web page
   <URL:http://gewis.win.tue.nl/~koen/conneg/>.

   The transparent content negotiation document set currently consists
   of three series of internet drafts.

     1. draft-ietf-http-negotiation-XX.txt

        `Transparent Content Negotiation in HTTP'

        Defines the core mechanism. Standards track.

     2. draft-ietf-http-rvsa-v10-XX.txt  (this document)

        `HTTP Remote Variant Selection Algorithm -- RVSA/1.0'

        Defines the remote variant selection algorithm version 1.0.
        Standards track.

     3. draft-ietf-http-feature-reg-XX.txt

        `Feature Tag Registration Procedures'

        Defines feature tag registration.  Best Current Practice track.

   An additional document about `the core feature set', which may
   later become an informational RFC, may also appear.  Currently,
   there are two internet drafts which discuss parts of what could be
   a core feature set: draft-mutz-http-attributes-XX.txt and
   draft-goland-http-headers-XX.txt

   Older versions of the text in documents 1 and 2 may be found in the
   draft-holtman-http-negotiation-XX.txt series of internet drafts.


TABLE OF CONTENTS

     1  Introduction
      1.1 Revision history

     2  Terminology and notation

     3  The remote variant selection algorithm
      3.1 Input
      3.2 Output
      3.3 Computing overall quality values
      3.4 Definite and speculative quality values
      3.5 Determining the result

     4  Use of the algorithm
      4.1 Using quality factors to rank preferences
      4.2 Construction of short requests
      4.2.1 Collapsing Accept header elements
      4.2.2 Omitting Accept headers
      4.2.3 Dynamically lengthening requests
      4.3 Differences between the local and the remote algorithm
      4.3.1 Avoiding major differences
      4.3.2 Working around minor differences

     5  Security and privacy considerations

     6  Acknowledgments

     7  References

     8  Authors' addresses


1  Introduction

   HTTP allows web site authors to put multiple versions (variants) of
   the same information under a single URL.  Transparent content
   negotiation [5] is a mechanism for automatically selecting the best
   variant when the URL is accessed.  A remote variant selection
   algorithm can be used by a server to choose a best variant on
   behalf of a negotiating user agent.  The use of a remote algorithm
   can speed up the transparent negotiation process by eliminating a
   request-response round trip.

   This document defines the remote variant selection algorithm with
   the version number 1.0.  The algorithm computes whether the Accept
   headers in the request contain sufficient information to allow a
   choice, and if so, which variant must be chosen.


1.1 Revision history

   Most text in section 3 of this draft was taken from section 11 of
   the draft-holtman-http-negotiation-04.txt internet draft.  Major
   changes are:

     - The scope of the `network negotiation algorithm' has been
       limited to variant selection by servers on behalf of the user
       agent only.  The algorithm has been renamed to `remote variant
       selection algorithm'.

   In the text and HTML versions with change marks, change bars or
   change marks only appear in section 3.  In the text version, all
   changes are marked, except changes in formatting and punctuation.
   In the HTML version, all changed words and symbols are typeset in
   bold text.  Deletions and changes in punctuation are not marked.

   Section 4 of this draft contains almost exclusively new material.


2  Terminology and notation

   This specification uses the terminology and notation of the HTTP
   transparent content negotiation specification [5].


3  The remote variant selection algorithm

   This section defines the remote variant selection algorithm with
   the version number 1.0.  To implement this definition, a server may
   run any algorithm which gives equal results.

     Note: According to [5], servers are always free to return a list
     response instead of running a remote algorithm.  Therefore,
     whenever a server may run a remote algorithm, it may also run a
     partial implementation of the algorithm, provided that the
     partial implementation always returns List_Response when it
     cannot compute the real result.


3.1 Input

   The algorithm is always run for a particular request on a
   particular transparently negotiable resource.  It takes the
   following information as input.

    1. The variant list of the resource, as present in the Alternates
       header of the resource.

    2. (Partial) Information about capabilities and preferences of the
       user agent for this particular request, as given in the Accept
       headers of the request.

   If a fallback variant description

       {"fallback.html"}

   is present in the Alternates header, the algorithm must interpret
   it as the variant description

       {"fallback.html" 0.00000000000000000001}

   The extremely low source quality value ensures that the fallback
   variant only gets chosen if all other options are exhausted.


3.2 Output

   As its output, the remote variant selection algorithm and will
   yield the appropriate action to be performed.  There are two
   possibilities:

      Choice_response

        The Accept headers contain sufficient information to make a
        choice on behalf of the user agent possible, and the best
        variant may be returned in a choice response.

      List_response

        The Accept headers do not contain sufficient information to
        make a choice on behalf of the user agent possible.  A list
        response must be returned, allowing the user agent to make the
        choice itself.


3.3 Computing overall quality values

   As a first step in the remote variant selection algorithm, the
   overall qualities of the individual variants in the list are
   computed.

   The overall quality Q of a variant is the value

      Q = qs * qt * qc * ql * qf

   where the factors qs, qt, qc, ql, and qf are determined as follows.

      qs Is the source quality factor in the variant description.

      qt The media type quality factor is 1 if there is no type
         attribute in the variant description, or if there is no
         Accept header in the request.  Otherwise, it is the quality
         assigned by the Accept header to the media type in the type
         attribute.

           Note: If a type is matched by none of the elements of an
           Accept header, the Accept header assigns the quality factor
           0 to that type.

      qc The charset quality factor is 1 if there is no charset
         attribute in the variant description, or if there is no
         Accept-Charset header in the request.  Otherwise, the charset
         quality factor is the quality assigned by the Accept-Charset
         header to the charset in the charset attribute, see section
         8.2 of [5].

      ql The language quality factor is 1 if there is no language
         attribute in the variant description, or if there is no
         Accept-Language header in the request.  Otherwise, the
         language quality factor is the highest quality factor
         assigned by the Accept-Language header to any one of the
         languages listed in the language attribute.

      qf The features quality factor is 1 if there is no features
         attribute in the variant description, or if there is no
         Accept-Features header in the request.  Otherwise, it is the
         quality degradation factor for the features attribute, see
         section 6.4 of [5].

   As an example, if a variant list contains the variant description

     {"paper.html.en" 0.7 {type text/html} {language fr}}

   and if the request contains the Accept headers

     Accept: text/html:q=1.0, */*:q=0.8
     Accept-Language: en;q=1.0, fr;q=0.5

   the remote variant selection algorithm will compute an overall
   quality for the variant as follows:

     {"paper.html.fr" 0.7 {type text/html} {language fr}}
                       |           |                 |
                       |           |                 |
                       V           V                 V
                      0.7   *     1.0        *      0.5  = 0.35

   With the above Accept headers, the complete variant list

     {"paper.html.en" 0.9 {type text/html} {language en}},
     {"paper.html.fr" 0.7 {type text/html} {language fr}},
     {"paper.ps.en"   1.0 {type application/postscript} {language en}}

   would yield the following computations:

                     qs  * qt  * qc  * ql  * qf  =  Q
                     ---   ---   ---   ---   ---   ----
      paper.html.en: 0.9 * 1.0 * 1.0 * 1.0 * 1.0 = 0.9
      paper.html.fr: 0.7 * 1.0 * 1.0 * 0.5 * 1.0 = 0.35
      paper.ps.en:   1.0 * 0.8 * 1.0 * 1.0 * 1.0 = 0.8


3.4 Definite and speculative quality values

   A computed overall quality value can be either definite or
   speculative.  An overall quality value is definite if it was
   computed without using any wildcard characters '*' in the Accept
   headers, and without the need to use the absence of a particular
   Accept header.  An overall quality value is speculative otherwise.

   As an example, in the previous section, the quality values of
   paper.html.en and paper.html.fr are definite, and the quality value
   of paper.ps.en is speculative because the type
   application/postscript was matched to the range */*.

   Definiteness can be defined more formally as follows.  An overall
   quality value Q is definite if the same quality value Q can be
   computed after the request message is changed in the following way:

       1. If an Accept, Accept-Charset, Accept-Language, or
          Accept-Features header is missing from the request, add
          this header with an empty field.

       2. Delete any media ranges containing a wildcard character '*'
          from the Accept header.  Delete any wildcard '*' from the
          Accept-Charset, Accept-Language, and Accept-Features
          headers.

   As another example, the overall quality factor for the variant

     {"blah.html" 1 {language en-gb} {features blebber [x y]}}

   is 1 and definite with the Accept headers

     Accept-Language: en-gb, fr
     Accept-Features: blebber, x, !y, *

   and

     Accept-Language: en, fr
     Accept-Features: blebber, x, *

   The overall quality factor is still 1, but speculative, with the
   Accept headers

     Accept-language: en-gb, fr
     Accept-Features: blebber, !y, *

   and

     Accept-Language: fr, *
     Accept-Features: blebber, x, !y, *


3.5 Determining the result

   The best variant, as determined by the remote variant selection
   algorithm, is the one variant with the highest overall quality
   value, or, if there are multiple variants which share the highest
   overall quality, the first variant in the list with this value.

   The end result of the remote variant selection algorithm is
   Choice_response if all of the following conditions are met

      a. the overall quality value of the best variant is greater
         than 0

      b. the overall quality value of the best variant is a definite
         quality value

      c. the negotiable resource and the best variant resource are
         neighbors.  This last condition exists for security reasons:
         it prevents some forms of spoofing, see section 14.2 of [5].

   In all other cases, the end result is List_response.

   The requirement for definiteness above affects the interpretation
   of Accept headers in a dramatic way.  For example, it causes the
   remote algorithm to interpret the header

     Accept: image/gif;q=0.9, */*;q=1.0

   as

     `I accept image/gif with a quality of 0.9, and assign quality
     factors up to 1.0 to other media types.  If this information is
     insufficient to make a choice on my behalf, do not make a choice
     but send the list of variants'.

   Without the requirement, the interpretation would have been

     `I accept image/gif with a quality of 0.9, and all other media
     types with a quality of 1.0'.


4  Use of the algorithm

   This section discusses how user agents can use the remote algorithm
   in an optimal way.  This section is not normative, it is included
   for informational purposes only.


4.1 Using quality factors to rank preferences

   Using quality factors, a user agent can not only rank the elements
   within a particular Accept header, it can also express precedence
   relations between the different Accept headers.  Consider for
   example the following variant list:

     {"paper.english" 1.0 {language en} {charset ISO-8859-1}},
     {"paper.greek"   1.0 {language el} {charset ISO-8859-7}}

   and suppose that the user prefers "el" over "en", while the user
   agent can render "ISO-8859-1" with a higher quality than
   "ISO-8859-7".  If the Accept headers are

     Accept-Language: gr, en;q=0.8
     Accept-Charset: ISO-8859-1, ISO-8859-7;q=0.6, *

   then the remote variant selection algorithm would choose the
   English variant, because this variant has the least overall quality
   degradation.  But if the Accept headers are

     Accept-Language: gr, en;q=0.8
     Accept-Charset: ISO-8859-1, ISO-8859-7;q=0.95, *

   then the algorithm would choose the Greek variant.  In general, the
   Accept header with the biggest spread between its quality factors
   gets the highest precedence.  If a user agent allows the user to
   set the quality factors for some headers, while other factors are
   hard-coded, it should use a low spread on the hard-coded factors
   and a high spread on the user-supplied factors, so that the user
   settings take precedence over the built-in settings.


4.2 Construction of short requests

   In a request on a transparently negotiated resource, a user agent
   need not send a very long Accept header, which lists all of its
   capabilities, to get optimal results.  For example, instead of
   sending

     Accept: image/gif;q=0.9, image/jpeg;q=0.8, image/png;q=1.0,
             image/tiff;q=0.5, image/ief;q=0.5, image/x-xbitmap;q=0.8,
             application/plugin1;q=1.0, application/plugin2;q=0.9

   the user agent can send

     Accept: image/gif;q=0.9, */*;q=1.0

   It can send this short header without running the risk of getting a
   choice response with, say, an inferior image/tiff variant.  For
   example, with the variant list

     {"x.gif" 1.0 {type image/gif}}, {"x.tiff" 1.0 {type image/tiff}},

   the remote algorithm will compute a definite overall quality of 0.9
   for x.gif and a speculative overall quality value of 1.0 for
   x.tiff.  As the best variant has a speculative quality value, the
   algorithm will not choose x.tiff, but return a list response, after
   which the selection algorithm of the user agent will correctly
   choose x.gif.  The end result is the same as if the long Accept
   header above had been sent.

   Thus, user agents can vary the length of the Accept headers to get
   an optimal tradeoff between the speed with which the first request
   is transmitted, and the chance that the remote algorithm has enough
   information to eliminate a second request.

4.2.1 Collapsing Accept header elements

   This section discusses how a long Accept header which lists all
   capabilities and preferences can be safely made shorter.  The
   remote variant selection algorithm is designed in such a way that
   it is always safe to shorten an Accept or Accept-Charset header by
   two taking two header elements `A;q=f' and `B;q=g' and replacing
   them by a single element `P;q=m' where P is a wildcard pattern that
   matches both A and B, and m is the maximum of f and g.  Some
   examples are

      text/html;q=1.0, text/plain;q=0.8       -->    text/*;q=1.0
      image/*;q=0.8, application/*;q=0.7      -->    */*;q=0.8

      iso-8859-5;q=1.0, unicode-1-1;q=0.8     -->    *;q=1.0

   Note that every `;q=1.0' above is optional, and can be omitted:

      iso-8859-7;q=0.6, *                     -->    *

   For Accept-Language, it is safe to collapse all language ranges
   with the same primary tag into a wildcard:

      en-us;q=0.9, en-gb;q=0.7, en;q=0.8, da  -->    *;q=0.9, da

   It is also safe to collapse a language range into a wildcard, or to
   replace it by a wildcard, if its primary tag appears only once:

      *;q=0.9, da                             -->    *

   Finally, in the Accept-Features header, every feature expression
   can be collapsed into a wildcard, or replaced by a wildcard:

      colordepth<=5, *                        -->    *

4.2.2 Omitting Accept headers

   According to the HTTP/1.1 specification [1], the complete absence
   of an Accept header from the request is equivalent to the presence
   of `Accept: */*'.  Thus, if the Accept header is collapsed to
   `Accept: */*', a user agent may omit it entirely.  An
   Accept-Charset, Accept-Language, or Accept-Features header which
   only contains `*' may also be omitted.

4.2.3 Dynamically lengthening requests

   In general, a user agent capable of transparent content negotiation
   can send short requests by default.  Some short Accept headers
   could be included for the benefit of existing servers which use
   HTTP/1.0 style negotiation (see section 4.2 of [5]).  An example is

      GET /paper HTTP/1.1
      Host: x.org
      User-Agent: WuxtaWeb/2.4
      Negotiate: 1.0
      Accept-Language: en, *;q=0.9

   If the Accept headers included in such a default request are not
   suitable as input to the remote variant selection algorithm, the
   user agent can disable the algorithm by sending `Negotiate: trans'
   instead of `Negotiate: 1.0'.

   If the user agent discovers, though the receipt of a list or choice
   response, that a particular origin server contains transparently
   negotiated resources, it could dynamically lengthen future requests
   to this server, for example to

      GET /paper/chapter1 HTTP/1.1
      Host: x.org
      User-Agent: WuxtaWeb/2.4
      Negotiate: 1.0
      Accept: text/html, application/postscript;q=0.8, */*
      Accept-Language: en, fr;q=0.5, *;q=0.9
      Accept-Features: tables, *

   This will increase the chance that the remote variant selection
   algorithm will have sufficient information to choose on behalf of
   the user agent, thereby optimizing the negotiation process.  A good
   strategy for dynamic extension would be to extend the headers with
   those media types, languages, charsets, and feature tags mentioned
   in the variant lists of past responses from the server.


4.3 Differences between the local and the remote algorithm

   A user agent can only optimize content negotiation though the use
   of a remote algorithm if its local algorithm will generally make
   the same choice.  If a user agent receives a choice response
   containing a variant X selected by the remote algorithm, while the
   local algorithm would have selected Y, the user agent has two
   options:

     1. Retrieve Y in a subsequent request. This is sub-optimal
        because it takes time.

     2. Display X anyway.  This is sub-optimal because it makes the
        end result of the negotiation process dependent on factors
        that can randomly change.  For the next request on the same
        resource, and intermediate proxy cache could return a list
        response, which would cause the local algorithm to choose and
        retrieve Y instead of X.  Compared to a stable
        representation, a representation which randomly switches
        between X and Y (say, the version with and without frames) has
        a very low subjective quality for most users.

   As both alternatives above are unattractive, a user agent should
   try to avoid the above situation altogether.  The sections below
   discuss how this can be done.

4.3.1 Avoiding major differences

   If the user agent enables the remote algorithm in this
   specification, it should generally use a local algorithm which
   closely resembles the remote algorithm.  The algorithm should for
   example also use multiplication to combine quality factors.  If the
   user agent combines quality factors by addition, it would be more
   advantageous to define a new remote variant selection algorithm,
   with a new major version number, for use by this agent.

4.3.2 Working around minor differences

   Even if a local algorithm uses multiplication to combine quality
   factors, it could use an extended quality formulae like

      Q = ( qs * qt * qc * ql * qf ) * q_adjust

   in order to account for special interdependencies between
   dimensions, which are due to limitations of the user agent.  For
   example, if the user agent, for some reason, cannot handle the
   iso-8859-7 charset when rendering text/plain documents, the
   q_adjust factor would be 0 when the text/plain - iso-8859-7
   combination is present in the variant description, and 1 otherwise.

   By selectively withholding information from the remote variant
   selection algorithm, the user agent can ensure that the remote
   algorithm will never make a choice if the local q_adjust is less
   than 1.  For example, to prevent the remote algorithm from ever
   returning a text/plain - iso-8859-7 choice response, the user agent
   should take care to never produce a request which exactly specifies
   the quality factors of both text/plain and iso-8859-7.  The
   omission of either factor from a request will cause the overall
   quality value of any text/plain - iso-8859-7 variant to be
   speculative, and variants with speculative quality values can never
   be returned in a choice response.

   In general, if the local q_adjust does not equal 1 for a particular
   combination X - Y - Z, then a remote choice can be prevented by
   always omitting at least one of the elements of the combination
   from the Accept headers, and adding a suitable wildcard pattern to
   match the omitted element, if such a pattern is not already
   present.


5  Security and privacy considerations

   This specification introduces no security and privacy
   considerations not already covered in [5].  See [5] for a
   discussion of privacy risks connected to the sending of Accept
   headers.


6  Acknowledgments

   Work on HTTP content negotiation has been done since at least 1993.
   The authors are unable to trace the origin of many of the ideas
   incorporated in this document.  This specification builds on an
   earlier incomplete specification of content negotiation recorded in
   [2].  Many members of the HTTP working group have contributed to
   the negotiation model in this specification.  The authors wish to
   thank the individuals who have commented on earlier versions of
   this document, including Brian Behlendorf, Daniel DuBois, Ted
   Hardie, Larry Masinter, and Roy T. Fielding.


7  References

   [1] R. Fielding, J. Gettys, J. C. Mogul, H. Frystyk, and
       T. Berners-Lee.  Hypertext Transfer Protocol -- HTTP/1.1.  RFC
       2068, HTTP Working Group, January, 1997.

   [2] Roy T. Fielding, Henrik Frystyk Nielsen, and Tim Berners-Lee.
       Hypertext Transfer Protocol -- HTTP/1.1.  Internet-Draft
       draft-ietf-http-v11-spec-01.txt, HTTP Working Group, January,
       1996.

   [3] T. Berners-Lee, R. Fielding, and H. Frystyk.  Hypertext
       Transfer Protocol -- HTTP/1.0.  RFC 1945.  MIT/LCS, UC Irvine,
       May 1996.

   [4] K. Holtman, A. Mutz.  Feature Tag Registration Procedures.
       Internet-Draft draft-ietf-http-feature-reg-00.txt, HTTP Working
       Group, October 30, 1996.

   [5] K. Holtman, A. Mutz.  Transparent Content Negotiation in HTTP.
       Internet-Draft draft-ietf-http-negotiation-00.txt, HTTP Working
       Group.


8  Authors' addresses

   Koen Holtman
   Technische Universiteit Eindhoven
   Postbus 513
   Kamer HG 6.57
   5600 MB Eindhoven (The Netherlands)
   Email: koen@win.tue.nl

   Andrew H. Mutz
   Hewlett-Packard Company
   1501 Page Mill Road 3U-3
   Palo Alto CA 94304, USA
   Fax +1 415 857 4691
   Email: mutz@hpl.hp.com


Expires: August 5, 1997