[Search] [txt|pdf|bibtex] [Tracker] [WG] [Email] [Nits]

Versions: 00 01 02 03 04 05 06 07 08 09                                 
EDIINT Functional Specification               July 1996

EDIINT Working Group                     Chuck Shih
Internet Draft                           Mats Jansson
Expires:  12/96                          Rik Drummond
                                         Lincoln Yarbrough

      Requirements for Inter-operable Internet EDI

Please direct comments to drummond@onramp.net.

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 made obsolete 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),
   unnari.oz.au (Pacific Rim), ds.internic.net (US East
   Coast), or ftp.isi.edu (US West Coast).

   Any questions, comments, and reports of defects or
   ambiguities in this specification may be sent to the
   mailing list for the EDIINT working group of the IETF,
   using the address <ietf-ediint@imc.org>. Requests to
   subscribe to the mailing list should be addressed to


   This memo is an informational document discussing the
   requirements  for inter-operable EDI, with sufficient
   background material to give an explanation for the EDI
   community of the Internet-related issues.

                      Table of Contents

1.0  INTRODUCTION                                          4

 1.1  THE AUDIENCE                                        4

2.0 THE INTERNET - A BRIEF HISTORY                         5

 2.1. THE INTERNET - MYTHS AND REALITY                    6

 2.2 INTERNET ROUTING                                     7

3.0 FUNCTIONAL REQUIREMENTS                                9

 3.1   INTRODUCTION AND DEFINITIONS                       9

 ENCRYPTION.                                              9
  3.2.1  INTRODUCTION AND DESCRIPTION                     9
  3.2.2  SYMMETRIC ENCRYPTION                            10
  3.2.4  NEEDS                                           11
  3.2.5  ISSUES                                          11

 3.3  KEY MANAGEMENT - SYMMETRIC KEYS                    13
  3.3.1 INTRODUCTION AND DESCRIPTION                     13
  3.3.2  NEEDS                                           15
  3.3.3  ISSUES                                          15
  3.3.4   RECOMMENDATION                                 15
  3.4.1 INTRODUCTION AND DESCRIPTION                     16
  3.4.2 PUBLIC KEYS                                      16
  3.4.3 NEEDS                                            16
  3.4.4 ISSUES                                           16
  3.4.5  RECOMMENDATIONS                                 16

 3.5  CONTENT INTEGRITY                                  17
  3.5.1  INTRODUCTION AND DESCRIPTION                    17
  3.5.2  NEEDS                                           18
  3.5.3 ISSUES                                           18
  3.5.4  RECOMMENDATIONS                                 18

  3.6.1  INTRODUCTION AND DESCRIPTION                    18
  3.6.2 NEEDS                                            19
  3.6.4  EDITOR'S RECOMMENDATIONS                        20

  3.7.1  INTRODUCTION AND DESCRIPTION                    20
  3.7.2  NEEDS                                           21
  3.7.3  RECOMMENDATIONS                                 21

  3.8.1 INTRODUCTION AND DESCRIPTION                     22
  3.8.2   GATEWAY FUNCTIONS                              22

 SERVICES                                                22
  3.9.1  INTRODUCTION AND DESCRIPTION                    22
  3.9.2 NEEDS                                            22
  3.9.3 ISSUES                                           23
  3.9.4  RECOMMENDATION                                  23

4.0 TRACKING AND ERROR HANDLING BASICS                    23

 4.1 INTRODUCTION                                        23

  4.2.1 NEED                                             24
  4.2.2 RECOMMENDATIONS                                  24

 AND SENT                                                24
  4.3.1 NEED                                             24
  4.3.2 RECOMMENDATIONS                                  25

  4.4.1 NEED                                             25
  4.4.2  RECOMMENDATIONS                                 25

 RECEIVER                                                25
  4.5.1 NEED                                             25
  4.5.2 RECOMMENDATIONS                                  25

 TRANSMISSIONS                                           25
  4.6.1  NEED                                            25
  4.6.2  RECOMMENDATIONS                                 26

  4.7.1  NEED                                            26


1.0  Introduction

   Electronic Data Interchange (EDI) is a set of protocols
   for conducting highly structured inter-organization
   exchanges, such as for making purchases or initiating
   loan requests.  The initial RFC1767 defined the method
   for packaging EDI X12 and UN/EDIFACT transaction sets in
   a MIME envelope.  However, several additional
   requirements for obtaining multi-vendor, inter-operable
   service, over and above how the EDI transactions are
   packaged, have come to light since the effort concluded.
   These currently revolve around security issues such as
   EDI transaction integrity, confidentiality and non-
   repudiation in various forms.  Additional requirements
   that mimic many of the heading fields found in X.435 EDI
   messages (e.g., Interchange Sender, Interchange
   Recipient, Interchange Control Reference, Communications
   Agreement ID, and Syntax Identifier) are also needed to
   support efficient exchanges by gateways and Value Added
   Networks.   Standards in these and other areas are
   necessary to ensure inter-operability between EDI
   packages over the Internet. Various technologies already
   exist for these additional features and the primary
   requirement is to review and select a common set of
   components for use by the EDI community when it sends
   EDI over the Internet.  In effect, the effort is to
   provide an EDI over the Internet Informational Document
   and an Applicability Statement Document.

   This document's current focus is on EDI MIME content
   transported using SMTP (Simple Mail Transport Protocol),
   the Internet's mail or messaging system.

   Traditional VAN connectivity is slow and expensive. The
   Internet promises lower cost usage and is more easily
   accessible than traditional methods of communications.
   The predominant problem with the use of  the Internet
   for EDI is interoperability between vendor products,
   specifically in the areas of integrity, confidentiality,
   signature, and non-repudiation. The EDIINT working
   group's focus is to recommend solutions for each of
   these areas using existing standards whenever possible.

1.1  The Audience
   The audience for this document consists of persons
   directly or indirectly involved in EDI communications
   decisions, companies trading EDI documents today or in
   the future, and  vendors developing and marketing EDI
   products. Also included in the audience for this
   document, are people providing services and consulting
   to the EDI community.

2.0 The Internet - A Brief History
   The Internet is a world-wide collection of computers,
   routers, and networks, connected together using the
   TCP/IP suite of protocols. The Internet itself is not a
   network, but a collection of networks. The Internet was
   designed to be decentralized, with no single authority
   needed to run it. All hosts on the Internet can
   communicate with one another as peers, and all of the
   communications protocols are "open" -- the standards are
   in the public domain, and the standardization process is
   open to anyone willing to put in the hard work to help
   define them.

   One consequence of this standards "openness" is that the
   Internet can accommodate many different kinds of
   machines (toasters for example). Its protocols -- the
   TCP/IP suite - have therefore become the de facto
   standards for heterogeneous computer networking.  At one
   level, the Internet is a physical collection of
   computers connected by common protocols. At another
   level though, the Internet can be thought of as a
   distributed medium, offering  some important advantages
   for doing EDI: for instance, the Internet has hundreds
   of thousands of connected global hosts, and tens of
   millions of users. The Internet offers a flat rate,
   volume and time-of-day independent pricing structure for
   data transmission. The Internet is highly redundant,
   offering  the ability to route data along alternate
   paths. The Internet's decentralized structure makes
   adding new hosts relatively easy -- it scales well and
   supports high bandwidth communications technologies.

2.1. The Internet - Myths and Reality

   The Internet had its beginnings in 1969 as an
   experimental  U.S. Defense Department network called
   ARPANET. The network was built to facilitate research on
   how to construct networks that could withstand outages
   to part of the network, but continue to function.
   Network reliability was a fundamental design point when
   developing the architecture and protocols associated
   with the Internet. From the premise that the network was
   inherently unreliable (parts of it could be destroyed at
   any moment) emerged a design that was both robust and
   reliable.  Early on, the networks comprising the
   Internet were primarily those from government agencies
   and educational institutions. Access to the Internet was
   pretty much available only to computer science
   researchers, and government employees and their

   In 1986, the National Science Foundation, in order to
   provide access to what was then a scarce resource, put
   together an initiative to link five super-computer
   centers together using the TCP/IP protocols. Two very
   important results of the NSFNET initiative were the
   upgrading of the Internet's infrastructure with ever
   more powerful processors and higher speed links, and
   expansion of access to a larger user community.  The
   1990's has seen the continual upgrading of the Internet
   infrastructure and its expansion to new constituencies
   outside the traditional government and university
   research community. Commercial interests are now the
   largest as well as the fastest growing segment of the

   Commercial Internet providers, such as Performance
   Systems International (PSI), and UUNET (the Alternet
   network), have emerged from the collection of
   intermediate-level networks that came into being as a
   result of the NSFNET initiative. The national long
   distance carriers such as MCI, AT&T, and Sprint all
   provide commercial Internet services. These commercial
   providers, called Internet Service Providers or ISPs for
   short, make available Internet connectivity and various
   other Internet services to their clients. The perception
   of the Internet as experimental, and primarily used by
   and for      educational and research activities is
   rooted in its past, and does not reflect today's
   situation. The growth in commercial access to the
   Internet, along with the growth of the ISPs, is
   radically changing the Internet's network composition.

   The design and architecture underlying the Internet has
   proven its robustness by scaling to unprecedented
   proportions. The Internet is reliable from several

   (1) Technologically, the TCP/IP suite of protocols and
   the architecture underlying the Internet are stable and

   (2) Product implementations based on the TCP/IP suite
   are also stable and mature.

   (3) Internet routing is dynamic, so packets sent through
   the Internet get to their destinations even if there are
   network outages along the way.

   (4) The commercial ISP administered portions of the
   Internet, provide essentially the same level of network
   reliability, availability, monitoring, throughput,
   implementation and support services as existing EDI
   Value Added Networks (VANS), but at a lower cost and
   with higher bandwidth.

   Although the Internet is reliable, low-cost, highly
   accessible, supports high bandwidth communications, and
   is technically mature, there are still some valid
   concerns relating to the use of the Internet for EDI.
   These concerns revolve primarily around security,
   message tracking, audit trails, and authorization. Some
   of the concerns, encryption for instance, are of a
   general nature and not specific to the Internet--
   encryption may be required regardless of what type of
   network is traversed. Other concerns, such as tracking,
   arise because they are required by EDI, and supported by
   existing EDI Value Added Networks.

2.2 Internet Routing
   By using a common network trace program called
   Traceroute, the route traversed by a packet from a
   source host to a destination host on the Internet may be
   followed. Tracing routes on the Internet yield some
   interesting characteristics. As expected, the routes
   traversed go through the networks administered by the
   ISPs of each of the trading partners. Each route
   consists of multiple nodes through each network. The
   route can vary but that is not the typical case. The IP
   packets are delivered reliably, and within a specified
   period of time. When a reputable commercial ISP is used,
   none of the nodes in the route are administered by
   government or educational entities.

   By looking at Internet network traces, we can conclude
   that the  Internet does a very  good job of getting
   packets from a source to destination. However, between
   the source and the destination, the packets are routed
   through many  intermediate nodes. It is at the
   intermediate nodes where anyone on one of the machines
   that handles the packets can re-assemble the packets
   that make up the EDI Interchange and can read it, copy
   it, alter it, or delete it. In the case where the EDI
   Interchange is carried using an e-mail transport (SMTP),
   the situation could arise where the message cannot be
   delivered to the final recipient  and the message must
   be stored at the intermediate node. Once again, the
   message is susceptible to any number of security

   The likelihood of security attacks, (especially if going
   through intermediate nodes administered by a quality
   ISP) are quite low, and practically speaking, may be
   quite difficult. Nonetheless, since the possibility
   exists, it is a concern - particularly if the packets
   contain high value or sensitive EDI, or electronic
   commerce transactions.

   The Internet is being singled out in this discussion
   because our focus is EDI over the Internet, but the
   possibility of security threats and administrative
   glitches exist even if the EDI Interchanges go through
   an EDI VAN. It really is a matter of management of the
   machines that make up the intermediate nodes in any
   network that is at issue.  There are measures that can
   be taken to defend against security concerns while an
   EDI interchange is in transit. These security measures
   are essential requirements when doing EDI over the
   Internet. Each of the security measures is described,
   issues are discussed, and recommended solutions given.

3.0 Functional Requirements

3.1   Introduction and Definitions
   The following sections describe the  functional and
   inter-operability requirements, as well as some of the
   practical considerations of sending and receiving EDI
   transactions on the Internet. It is assumed that the
   reader is generally familiar with EDI.

3.2  Standard Encryption Algorithms and World-Wide
3.2.1  Introduction and Description
   The goal of encryption is to turn otherwise readable
   text into something that cannot be read, and therefore
   understood. By making the text unintelligible,
   encryption discourages anyone from reading or copying
   the EDI Interchange while it is in transit between the
   trading partners. Encryption conveys confidentiality to
   the EDI Interchange. Traffic analysis is always
   possible, since many times, header information is not
   encrypted. (Traffic analysis is the analysis of header
   information in order to derive useful information from

   Encryption is based on two components: an algorithm and
   a key. An algorithm is a mathematical transformation
   that takes plain-text or other intelligible information
   and changes it into unintelligible cipher text. The
   inverse mathematical transformation, back to the
   original from the cipher text is also done, and  is
   called decryption. In order to encrypt the plain text, a
   key is used as input in conjunction with an encryption
   algorithm. An algorithm can use one of any of a large
   number of possible keys. The number of possible keys
   each algorithm can support depends on the number of bits
   in the key. For instance, if the key length is 40, then
   2 to the n, where n is the number of bits in the key,
   results in 1,000,000,000,000 possible key combinations,
   with each different key causing the algorithm to produce
   slightly different cipher output.

   An encryption algorithm is considered secure if its
   security is dependent only on the length of its key.
   Security cannot be dependent on the secrecy of the
   algorithm, the inaccessibility of the cipher or plain
   text, or anything else--except the key length.  If  this
   were true about a particular algorithm, then the most
   efficient -- and only -- attack on that algorithm is a
   brute-force attack, whereby all  key combinations must
   be tried in order to find the one correct key (this is
   true for symmetric encryption algorithms, asymmetric
   algorithms work a little differently, and are explained
   in greater detail later).  So, by  specifying a long
   enough key length n, even a brute-force attack on a
   secure algorithm can be made completely non-feasible.

3.2.2  Symmetric Encryption
   Encryption algorithms whereby two trading partners must
   use the identical key to encrypt and decrypt the EDI
   Interchange are called symmetric encryption algorithms.
   Put another way, if an EDI Interchange is encrypted with
   one key, it cannot be decrypted with a different key.
   The key used in most symmetric encryption algorithms is
   just a random bit string, n  bits long. These keys are
   often generated from random data derived from the source

   The use of symmetric encryption simplifies the
   encryption process, each trading partner does not need
   to develop and exchange secret encryption algorithms
   with one another (which incidentally would be a near
   impossible task). Instead, each trading partner can use
   the same encryption algorithm, and only exchange the
   shared, secret key.

   There are drawbacks however with symmetric encryption
   schemes; a shared secret key must be agreed upon by both
   parties. If a trading partner has n trading
   relationships then n secret keys must be  maintained,
   one for each trading partner. Symmetric encryption
   schemes also have the problem that origin or destination
   authenticity (non-repudiation of origin, delivery and
   receipt) cannot be proved. Since both parties share the
   secret encryption key, any EDI Interchange encrypted
   with a symmetric key, could have been sent by either of
   the trading partners. By using what is called public key
   cryptography, which makes use of asymmetric encryption
   algorithms, the non-repudiation of origin, delivery and
   non-repudiation of receipt issues can be solved.

3.2.3  Asymmetric Encryption - Public-key Cryptography
   Public-key cryptography is based on the concept of a key
   pair. Each half of the pair (one key) can encrypt
   information that only the other half (one key) can
   decrypt. The key pair is designated and associated to
   one, and only one, trading partner. One part of the key
   pair (the private key) is only known by the designated
   trading partner; the other part of the key pair (the
   public key) is published widely but is still
   associated with the designated trading partner.

   The keys are used in different ways for confidentiality
   and digital signature. Both confidentiality and
   signature depend on each entity having a key pair that
   is identified only with them and each party keeping one
   pair of their key pair secret from all others.

   Signature  works as follows: Trading Partner A uses her
   secret key to encrypt part of a message, then sends the
   encrypted message to Trading Partner B. B gets Trading
   partner A's public key (anyone may get it) and attempts
   to decrypt the encrypted part of Trading partner A's
   message. If it decrypts, then Trading Partner B knows it
   is from A--because only A's public key can decrypt a
   message encrypted with  A's private key, and A is the
   only one who knows her private key.

   Confidentiality applies the asymmetric key pair, but in
   a different manner than signature. If Trading partner A
   wishes to send a confidential message to Trading Partner
   B she would apply the key pair as follows: Trading
   partner A would retrieve Trading partner B's public key,
   and encrypt the message with it. When Trading Partner B
   receives the message, she would decrypt the message
   with her private key. Only her private key can decrypt
   information that was encrypted with her public key. In
   other words, anything encrypted with B's public key can
   only be decrypted with B's private key.

   Since public-key encryption algorithms are considerably
   slower than their symmetric key cousins, they are
   generally not used to do the actual  encryption of what
   could be large EDI Interchanges. For instance, RSA Data
   Securities, Inc. estimates that software encryption
   using DES (a symmetric key algorithm) is 100 times
   faster than software encryption using RSA (a public-key
   encryption algorithm from RSA Data Securities, Inc.).
   Hardware encryption using DES is anywhere from 1,000 to
   10,000 times faster than hardware encryption using the
   RSA asymmetric encryption algorithm.  Instead of being
   used for bulk encryption, public-key encryption
   algorithms are used to encrypt symmetric encryption
   keys.  They are also used as an efficient means of
   exchanging and managing symmetric encryption keys.

3.2.4  Needs
   In order to provide confidentiality for EDI Interchanges
   on the Internet, a standard encryption algorithm(s) and
   key length(s) must be specified. For inter-operability
   to occur between two trading partners, the encryption
   algorithm and key lengths must be agreed upon either
   before hand or within an individual transaction.

3.2.5  Issues
   * When choosing an encryption algorithm, the following
   criteria should be considered; how secure the algorithm
   is;  how fast implementations of the algorithm are; the
   availability of the algorithms for international as well
   as domestic use; the availability of APIs and tool kits
   in order to implement the algorithms; and the frequency
   of the use of  the algorithm in existing

   * Sufficient key lengths must be chosen with regard to
   the value of the EDI Interchange so that brute-force
   attacks are not worth the time or effort compared to the
   value of the Interchange.

3.2.6  Some  Recommended Encryption Algorithms
   DES: The most widely used commercial encryption
   algorithm is DES. It is widely used in the banking
   industry for Electronic Funds Transfer (EFT). DES is
   also a U.S. government encryption standard. DES is in
   the public domain, which means anyone can implement the
   algorithm, including  those in the international
   community. DES was designed for, and is used for bulk
   encryption of  data. DES is prohibited by the U.S.
   government for export.

   The DES algorithm has been analyzed by cryptographers
   since the mid-1970s, and is considered secure: in other
   words, the security of DES is completely dependent on
   the length of its key. DES specifies a 56 bit key, so 2
   to the 56th or 10 to the 16th keys are possible.  A
   brute force attack, which means trying every single key
   to decrypt 8 bytes of known cipher-text into its
   corresponding 8 bytes of known plain-text is the best
   attack on the algorithm.

   The amount of time and money required to mount a
   successful brute force attack varies with the processing
   power used - and how lucky the attacker may be in
   generating a key that is close to the one used to
   encrypt the original EDI Interchange. Some estimates
   which have been put forth claim that a $1 million dollar
   hardware based, brute-force attack on DES would take 3.6
   hours to recover the DES key.  A corresponding  $1
   million dollar software based brute-force attack on DES
   would however take 3 years.  As the price/performance of
   processors decrease, a 56 bit key becomes less and less
   adequate in protecting EDI Interchanges of extreme
   value. Triple-DES, an algorithm with longer key length
   (discussed below) should be used in such cases.

   Triple-DES is a variant on DES that encrypts the EDI
   Interchange 3 times, with 2 independent 56 bit keys,
   giving it an effective key length of 112 bits. This
   makes a brute-force attack on Triple-DES not feasible.
   DES and Triple-DES actually can be implemented in 3
   different modes. It is recommended that DES and Triple-
   DES be used in Cipher Block Chaining (CBC) mode, which
   gives added protection by making each cipher-text block
   depend on each other, so changes in the cipher-text can
   be detected.

   RC2 and RC4: RC2 and RC4  are proprietary symmetric
   algorithms of  RSA Data Security. RC2 and RC4 unlike DES
   are variable-key length algorithms. By specifying
   differing key lengths, RC2 and RC4 can be configured to
   provide greater or lesser security.  RC2 and RC4 are
   alternatives to DES and have special export status,
   whereby 40 bit versions of RC2 and RC4, and 56 bit
   versions for foreign subsidiaries and overseas offices
   of U.S. companies have expedited export approval from
   the U.S. government. RSA claims that RC2 and RC4 are
   faster than DES when implemented in software. Several e-
   mail products such as Lotus Notes and Apple's Open
   Collaboration Environment make use of these algorithms.
   It is recommended that a key length of at least 128 bits
   be used when using  RC2 and RC4. A key length of 128
   bits would  make a brute-force attack impossible now and
   for the foreseeable future.

   IDEA: The International Data Encryption Algorithm was
   published in 1991. The symmetric algorithm is an
   iterated block cipher with a 64-bit block size and a 128-
   bit key size. The key length of IDEA is over twice that
   of DES and is longer than triple-DES. The IDEA algorithm
   is patented in both the United States and abroad. The
   IDEA algorithm in CBC mode is used by PGP (Pretty Good
   Privacy - a popular  electronic mail security program)
   for encryption. Individual users of PGP have a royalty
   free license to use the IDEA algorithm.

   There are many encryption algorithms that are secure and
   can provide confidentiality for an EDI Interchange. For
   interchanges that carry less value, 40-bit RC2 or 56-bit
   DES are probably adequate. For more valuable
   interchanges, use of Triple-DES, IDEA, or longer key
   length RC2 or RC4 is recommended.

   DES is currently in wide-spread use, and Triple-DES is
   projected to be in wide-spread use, as the 56-bit DES
   key limitation becomes less and less adequate.  The DES
   algorithm is available  for implementation outside the
   U.S., and the DES algorithm has been studied for a long
   time, and has been found to be secure.  RC2 and RC4 are
   useful because they allow the security of  the
   encryption - the key length specification, to be
   configurable.  (the RC2 and RC4  algorithms are
   proprietary, they can be exported outside the U.S. IDEA
   is a newer algorithm and has not been studied as much as
   DES. It has an adequate key-length, though it is not
   configurable. Indications are that IDEA is a secure
   algorithm and its use in PGP makes it the most widely
   used encryption algorithm for Internet electronic mail.

3.3  Key Management - Symmetric Keys
3.3.1 Introduction and Description
   The use of symmetric encryption is based on a shared
   secret. Two trading partners using a symmetric
   encryption algorithm must be able to do the following;
   generate a random symmetric key and agree upon its use;
   securely exchange the symmetric key with one another;
   set up a process to invalidate a symmetric key that has
   been compromised or needs changing. Each trading partner
   would then need to do this with each and every one of
   their trading partners. Management and distribution of
   symmetric keys can become an  insecure and cumbersome

   Pure symmetric key management schemes also have the
   problem that origin authenticity cannot be proved. Since
   two parties share a secret encryption key, any EDI
   Interchange encrypted with a symmetric key, could have
   been sent by either of the trading partners who have
   knowledge of the key.

   Using public-key cryptography, the management of
   symmetric keys is simplified and made more secure.
   Trading partners do not need to agree on secret
   symmetric keys as part of the trading partner
   relationship. Public-key cryptography also solves the
   origin authenticity problem that pure symmetric key
   management schemes have.

   By generating a unique symmetric encryption key for each
   EDI Interchange, and using public key cryptography,
   trading partners no longer need to secretly agree on a
   shared symmetric key in the trading partner
   relationship. A symmetric key can be randomly generated
   by the software for each EDI Interchange between trading
   partners.  Since a unique symmetric key is generated for
   each EDI Interchange, key maintenance is not needed.
   Trading partners do not need to invalidate compromised
   or expired keys. Each symmetric key is  used only one

   Additional security is also realized using  the method
   described above;  in the unlikely event that one of the
   symmetric keys is compromised, only one EDI transaction
   is affected, and not every transaction in the trading
   partner relationship.  Public-key encryption also
   provides a secure way of distributing symmetric keys
   between trading partners. Since only the receiving
   trading partner has knowledge of her private asymmetric
   key, she is the only one that can decrypt the symmetric
   key encrypted with her public asymmetric key - and is
   thus the only one who can use the symmetric key to
   decrypt the EDI Interchange.

   To impart confidentiality to an EDI Interchange which
   trading partner ABC sends to trading partner XYZ, the
   following steps would be performed:

          1) The EDI Translator outputs the EDI Interchange.

          2) A random symmetric key of the specified  length
          is generated.

          3) The EDI Interchange is encrypted using the
          randomly generated symmetric key with the chosen
          encryption algorithm.

          4) The random symmetric key is then encrypted
          using  5) XYZ's, the destination's, public
          asymmetric key.

          The encrypted symmetric key and encrypted EDI
          Interchange are then enveloped and sent to the
          trading partner.

   On the receiving side, the following steps would be

          1) The symmetric key is decrypted using XYZ's
          private asymmetric key.

          2) The decrypted symmetric key is then used to
          decrypt the EDI Interchange.

          3) The decrypted EDI Interchange is then routed to
          the EDI translator.

3.3.2  Needs
   A method to manage the symmetric encryption keys used in
   encrypting EDI Interchanges. The method should simplify
   the generation, maintenance, and distribution of the
   symmetric encryption keys. The method should also
   provide a secure channel for distributing the symmetric
   encryption keys between trading partners.

3.3.3  Issues
          * Choose a public-key encryption algorithm that
          facilitates key management of the symmetric
          encryption keys. The symmetric encryption keys
          should be generated on the fly for each EDI

          * When choosing a public-key encryption algorithm,
          the following criteria should be considered; how
          secure the algorithm is;  how fast implementations
          of the algorithm are; the availability of the
          algorithms for international as well as domestic
          use; the availability of APIs and tool kits in
          order to implement the algorithms; and the
          frequency  of the use of  the algorithm in
          existing implementations.

          * Sufficient key lengths must be chosen with
          regard to the value of the EDI Interchange so that
          brute-force attacks are not worth the time or
          effort compared to the value of the Interchange.

3.3.4   Recommendation
   1) RSA is a public-key cryptography algorithm that has
   become a de facto standard in its use for key
   management. Its use is recommended in managing and
   distributing symmetric encryption keys when doing EDI
   over the Internet.  The RSA public-key algorithm also
   has the advantage that it can be used freely outside the

   2) The mathematics of RSA are complicated, but is based
   on the difficulty in factoring large prime numbers. A
   public-key is generated by multiplying two large prime
   numbers together, deriving the private key from the
   public key involves factoring the large prime number. If
   the prime is large enough, this becomes an impossible
   task. The RSA key length is configurable, and is
   recommended to be at least 512 bits (154 digits long),
   and preferably 1024 bits (or 308 digits long) or longer.

3.4  Key Management - Public and Private Keys
3.4.1 Introduction and Description
   The use of public-key cryptography to simplify the
   management of the symmetric encryption keys presents the
   user with two problems; protecting the private key, and
   binding a trading partner's identity to his public key.
   Most likely the user will never know what his private
   key is. The software will generate a random private key,
   encrypt it, and store it in a file or database. The
   private key is accessed indirectly by the user through
   access to the software. User access to the software is
   generally controlled by a password, pass-phrase, and/or
   certain access rights. These are internal security
   policies, and are company specific.  It is important to
   control the access to the private key since any
   unauthorized access can lead eventually to the
   revocation of the corresponding public key.

3.4.2 Public Keys
   A public key is used by a originating  trading partner
   to encrypt a symmetric key, and as we'll discuss later,
   by a receiving trading partner to do authentication and
   non-repudiation of origin and receipt. Trading partners
   exchange public keys using  a  public key certificate.
   A public key certificate is defined in the X.509
   standards, and is a binding of  an entity's
   distinguished name (a formal way for identifying someone
   or something in the X.500 world, in our case it would be
   a trading partner) to a public key. A certificate also
   contains the digital signature of the issuer of the
   certificate, the identity of the issuer of the
   certificate, and an issuer specific serial number, a
   validity period for the certificate, and information to
   verify the issuer's digital signature. Certificate
   issuers are called Certification Authorities, and are
   trusted by both trading partners. In essence, a
   certificate is a digitally notarized binding of trading
   partner to public key.

3.4.3 Needs
   Adoption of a trust model and use of Certification
   Authorities for verifying public key certificates.

3.4.4 Issues
   Lack of Certification Authorities. To date only Verisign
   has stepped forward as a Certification Authority.  The
   U.S. Postal Service is interested in becoming a
   Certification Authority but is in the process of
   implementing the services.

3.4.5  Recommendations
   Since there already exists a trust relationship between
   EDI trading partners, until Certification Authorities
   become more prevalent, and the formats and protocols for
   interacting with Certification Authorities become
   standardized, it is recommended that the trading
   partners self-certify each other and exchange public
   keys as part of the trading partner relationship. The
   self-certified certificate should still be exchanged
   between trading partners in X.509 v3 format, so
   migration to exchange of notarized certificates is made

   Since the  formats and protocols for use in registering,
   requesting, and revoking certificates have not been
   standardized, the use of PCKS #10 and S/MIME is one
   recommendation. Also, PGP with its circle of trust is
   another valid option.  We recommend that S/MIME be
   tested for interoperability first, and that PGP v3.0 be
   tested later. Implementation of the standards for
   Certification Authority interactions being developed by
   the IETF-pkix  (public-key infrastructure X.509) work
   group should eventually be adopted.

3.5  Content Integrity
3.5.1  Introduction and Description
   Encryption guarantees the confidentiality of an EDI
   Interchange. Content integrity guarantees that the
   receiving trading partner gets  the EDI Interchange in
   its originally sent state. Content integrity assures
   that no modifications - additions, deletions, or changes
   - have been made to the EDI Interchange when it is in
   transit between trading partners.

   Content integrity is achieved if the sender includes
   with the EDI Interchange, an integrity control value.
   This value can be computed by using an appropriate
   cryptographic algorithm to 'fingerprint' the EDI
   Interchange. These cryptographic algorithms are called
   one-way hash functions or message integrity checks.
   Similar to encryption, a one-way hash function turns the
   EDI intelligible plain-text into something

   Unlike encryption algorithms however, one-way hash
   functions can't be decrypted.  One-way hash functions
   are constructed so the probability is infinitely small
   that some arbitrary length piece of plain-text can be
   hashed to a particular value, or that any two pieces of
   plain-text can be hashed to the same value. One-way hash
   values are usually  from 112 to 160 bits long. The
   longer the hash value, the more secure it is.

   One-way hash functions don't require a key, and the
   algorithm used must be agreed upon by the trading
   partners. To insure content integrity, the sending
   trading partner needs to calculate a one-way hash value
   of the EDI Interchange.  This value is unique and
   'fingerprints' the EDI Interchange. The sending trading
   partner sends the hash value along with the EDI
   Interchange. The receiving trading partner using the
   same one-way hash function calculates the hash value for
   the received EDI Interchange. If the received hash value
   matches the calculated hash value, then the receiving
   trading partner knows that the EDI Interchange has not
   been tampered with.

3.5.2  Needs
   Choice of a one-way hash algorithm to calculate the hash
   value required  to insure content integrity.

3.5.3 Issues
   The one-way hash algorithm should  be secure, publicly
   available, and should produce hash values of at least
   128 bits.

3.5.4  Recommendations
   MD5 is a one-way hash function that is publicly
   available, produces a 128 bit hash value called a
   Message Digest. It is widely used or recommended by most
   e-mail security programs and specifications, such as
   PEM, PGP, and S/MIME.

3.6  Authentication and Non-Repudiation of Origin
3.6.1  Introduction and Description
   Encryption guarantees confidentiality. Applying a one-
   way hash function guarantees content integrity. Both
   authentication and non-repudiation of origin guarantee
   the identity of the sender of the EDI Interchange. Non-
   repudiation of origin, identifies the original sender,
   and is the same as authentication when the EDI
   Interchange is sent point-to-point, i.e. when there is
   no forwarding involved. Authentication and non-
   repudiation of origin discourages any spoofing attacks
   that may occur while the EDI Interchange is in transit
   between the trading partners.

   Both authentication and non-repudiation of origin are
   accomplished using digital signatures. A digital
   signature is another application of public-key
   cryptography, and is explained in more detail in the
   following paragraphs.

   Up to this point, a receiving trading partner's public-
   key has been used to encrypt a symmetric key, and this
   symmetric key could  only be decrypted by the receiving
   trading partner's private key.  However the roles of the
   private and public keys can be reversed, so that you
   encrypt with the private key, and decrypt with the
   public key. Again the keys are reciprocal, if encryption
   is done with the private key, decryption can only be
   done with the public key.

   Since only trading partner ABC knows her own private-
   key, only trading partner ABC can encrypt something with
   that private-key. Encryption with a private key
   therefore has the effect of uniquely  identifying the
   person or entity doing the encryption. It is in effect,
   a digital signature. Since ABC's public-key is known to
   all his trading partners, they can all decrypt something
   encrypted with ABC's private-key. Decryption using ABC's
   public-key therefore has the effect of authenticating
   ABC as the trading partner that did the encryption, and
   it in effect verifies ABC's digital signature. ABC also
   cannot say that he did not do the encryption, since he
   is the only one with knowledge of his private key.  In
   this way, non-repudiation of origin is achieved.

   So what should a trading partner sign or encrypt with
   his private-key to guarantee authentication and non-
   repudiation of origin? Remember, public-key encryption
   algorithms are not meant to encrypt something very
   large, they are too slow for that. The symmetric key is
   encrypted with a public-key, and encrypting this with a
   private-key is not recommended, as this would allow
   other than the authorized recipient to decrypt the EDI
   Interchange. Since a one-way hash value is pretty small,
   usually only between 112-160 bits long, it is a natural
   choice for what can be digitally signed. If  the message
   integrity value was signed with a private key, then not
   only could authentication and non-repudiation of origin
   be guaranteed, but message integrity as well.

3.6.2 Needs
   Choice of a digital signature algorithm.

   3.6.3  Issues

   When choosing a digital signature algorithm, the
   following criteria should be considered; how secure the
   algorithm is;  how fast implementations of the algorithm
   are; the availability of the algorithms for
   international as well as domestic use; the availability
   of APIs and tool kits in order to implement the
   algorithms; and the frequency  of the use of  the
   algorithm in existing implementations.

   Sufficient key lengths must be chosen with regard to the
   value of the EDI Interchange so that brute-force attacks
   are not worth the time or effort compared to the value
   of the Interchange.

3.6.4  Editor's Recommendations
   In addition to using RSA to encrypt keys, it is
   recommended that RSA also be used for digital
   signatures. Again, the RSA algorithm  has the advantage
   that it can be used freely outside the U.S.

3.7  Section: Signed Receipt or Non Repudiation of Receipt
3.7.1  Introduction and Description
   The signed receipt (or alternately Non Repudiation of
   Receipt), is a receipt acknowledgment sent by the
   receiving trading partner to the sending trading partner.
   The receipt is used to solve the following issues when
   doing EDI over the Internet:

          * Lack of mailbox delivery notification within the
          Internet standards, though these are currently
          being addressed within RFCs 1891-1894.

          * Provide the equivalent of VAN mailbox delivery

          * Provide the equivalent of VAN mailbox pick-up

          * Provide the equivalent of VAN mailbox

          * Detect the situation where EDI Interchanges are
          maliciously deleted, or are not delivered by the

   Receipt of a signed receipt is an implicit
   acknowledgment of successful mailbox delivery. It is
   also an explicit acknowledgment that the Interchange was
   retrieved from the mailbox--pick-up notification. By
   having the receiver sign the  receipt, it authenticates
   that the intended receiver picked up the EDI Interchange-
   -mailbox authentication--and that the intended receiver
   verified the integrity of the EDI Interchange, and the
   identity of the sender. By returning the original one-
   way hash value (or original message) back in the  signed
   receipt, the sender can reconcile the acknowledged EDI
   Interchange with what was sent.

3.7.2  Needs
   Define the format and protocol for the signed  receipt
   so that it provides the following:

          * Implicit acknowledgment of  mailbox delivery of
          the EDI Interchange to the recipient.

          * Explicit acknowledgment  that the receiver has
          authenticated the sender and verified the
          integrity of the sent EDI Interchange.

          * Guarantees  non-repudiation of receipt when  the
          receipt acknowledgment is digitally signed by the
          receiving trading partner.

          * Provide  information in the signed  receipt so
          it can be used for tracking, logging, and
          reconciliation purposes.

   The re-transmission timer, and retry count  to detect
   lost Interchanges should be recommended, but needs to be
   configurable. Additionally, it should be specified what
   the receiver should do when it receives duplicate
   Interchanges, and what the sender should do when it
   receives duplicate receipt acknowledgments. The signed
   receipt should be applicable to more than just EDI.

3.7.3  Recommendations
   The syntax for a signed receipt should not be specific
   to EDI, since many of the uses of a signed receipt can
   be broadly applied to other MIME encapsulated objects.
   It is recommended that the results of the IETF receipt
   working group be adopted for use as a signed receipt.
   The receipt working group has published an Internet
   draft--draft-ietf-receipt-mdn-01--which can be obtained
   off of the IETF World Wide Web site. The EDIINT working
   group is working with the receipt working group to
   specify additional disposition-field values, as well as
   specification of  how the MDN (message disposition
   notification) should be used within an EDI environment.
   Specifically, within an EDI environment, requests for
   message disposition requests  cannot be silently
   ignored. In addition, if non-repudiation of receipt is
   agreed to by the trading partners, the original message
   sent must be returned in the third component of the
   message disposition notification--digitally  signed by
   the receiver of the EDI Interchange. The time-out and
   retry values for the message disposition notification
   should be recommended, but needs to be configurable.
   Duplicates should be checked by the UA and discarded.
   Until the receipt Internet draft is complete, a
   combination of RFC1847 and draft-ietf-receipts-mdn
   should be used to implement this functionality

3.8  EDI Object Boundaries and Transaction Privacy
3.8.1 Introduction and Description
   The specification by this work group applies to the EDI
   Interchange level, and not the group  or document level.
   Any security services, packaging, transport, or non-
   repudiation services are assumed to be applied to an EDI
   Interchange. Unlike the X12.58 and UN/EDIFACT 9735-5 and
   9735-6 security standards, the security services cannot
   be applied at a group or document level. The purpose of
   the specification is to move these services out of the
   translator, and into the "communications" subsystem. The
   "communications" subsystem should know as little about
   the structure of the EDI data as possible.

   The entire EDI Interchange including the enveloping
   headers (ISA/IEA or UNA/UNB/UNZ) are also encrypted.
   Since the routing of the EDI Interchange is through the
   Internet, and not a VAN, the sender/receiver ids are not
   used in mailbox routing, so the EDI envelops can be
   encrypted when sending EDI over the Internet.

   Only one EDI Interchange is sent per e-mail message.

3.8.2   Gateway Functions
   Situations exist whereby a VAN, or internal gateway, in
   order to route an EDI Interchange received on the
   Internet, will need to be able to access the information
   in the EDI envelope. The enveloping information as well
   as other useful gateway information may need to be
   copied and sent as a separate body part. It is proposed
   that additional fields be specified in RFC 1767 to
   accommodate EDI specific gateway routing requirements,
   and this be sent as a separate body part from the
   encrypted EDI Interchange.

3.9  Syntax and Protocol for Specifying Cryptographic
3.9.1  Introduction and Description
   Once cryptographic services are applied to EDI
   Interchanges, then the formats and protocols must be
   specified on how the cryptographic information is
   conveyed during  the EDI message exchange. Encryption
   algorithm information, one-way hash algorithm
   information, symmetric keys, initialization vectors, one-
   way hash values, public-key certificates, need to be
   enveloped and sent along with the EDI Interchange.

3.9.2 Needs
   Syntax and protocol for specifying EDI Interchanges that
   have had cryptography applied to it. Choose an existing
   standard and don't reinvent one.

3.9.3 Issues
   The syntax should be transport independent  so it can be
   used with different Internet transports. The standard
   should have broad support, and implementations should be
   available. Finally it should be international in nature.

3.9.4  Recommendation
   The  IETF EDIINT working group has put together a matrix
   comparing many of the different ways that EDI with
   cryptography applied to it can be transmitted. The use
   of S/MIME and PGP/MIME (version 3.0 with the elkins
   draft)  are both viable alternatives. Each has its
   strengths and weaknesses as the comparison matrix brings

   The S/MIME specification allows signed, and encrypted
   and signed to be distinguished. The signatories in an
   S/MIME encrypted and signed  message can be
   distinguished, which in certain EDI and electronic
   commerce situations is not acceptable. S/MIME specifies
   40 bit RC2 as the default encryption algorithm and key
   length. In some applications neither this default
   algorithm or key length are acceptable. S/MIME can
   accommodate other security algorithms and key lengths
   such as those recommended  in section 3.3.2 however.

   PGP/MIME supports a set profile of security algorithms
   and some user configurable key lengths. PGP/MIME  does
   not have the signatory problem as described above for
   S/MIME. However, PGP/MIME does not give the user as much
   flexibility in choosing algorithms and key lengths,
   although the security profile used by PGP/MIME is more
   than adequate to insure confidentiality, non-repudiation
   of origin, and message integrity.

4.0 Tracking and Error Handling Basics

4.1 Introduction
   It's important to recognize that traditional EDI via
   Value Added Networks have some inherent tracking
   mechanisms, that we have to either duplicate in Internet
   EDI, or decide we don't need.  In Internet EDI, there is
   no third party to call to track a transmission after it
   has left one company, and before it has been received by
   the second company.  Also, the move from VAN based EDI
   to Internet EDI changes the connectivity in other ways.
   From a more batch oriented store-and-forward technology
   to a more event driven routing technology.

   Aside from the communications between companies,
   "tracking" touches many other points within the trading
   companies.  This is where the use of 997 functional
   acknowledgments come in, the EDIFACT CONTRL message, and
   the common translator tracking of sequential group
   control numbers.  All of this needs to be considered in
   Internet EDI tracking.  In addition, some recent
   developments within S/MIME warrant some analysis--
   "positive acknowledgment", which refers to mail response
   not just if the delivery failed, but also if it

   What tracking information do we really need, and where
   does the UA have a role in providing it?

          1) Transmission successfully translated from
          internal format to EDI standard format

          2) Transmission successfully encrypted and sent
          (The equivalence of transmission successfully
          forwarded to receiver's VAN mailbox.)

          3) Transmission successfully received by the
          intended receiver and successfully decrypted (The
          equivalence of a VAN acknowledgment that sent
          transmission has been picked up by the receiver.)

          4) Transmission successfully translated by the
          receiver (the EDI Interchange was determined to be
          syntactically correct.)

          5) Detection and recovery of delayed or lost

          6) Detection and handling of duplicate

          7) Detection and handling of out-of-sequence

   The question of what the UA needs to track as compared
   to what the EDI translator tracks is addressed in the
   following sections.

   Needs, issues and recommendations will be discussed.

4.2 Section: Transmission successfully translated from
internal format to standard EDI format
4.2.1 Need
   There needs to be a facility by which a sender can
   assure that the EDI transmission was correctly
   translated and prepared for outbound transmission.

4.2.2 Recommendations
   This is standard functionality for most if not all EDI
   translators.  This should NOT be required functionality
   in the UA.

4.3 Section: Transmission successfully encrypted, signed and
4.3.1 Need
   There needs to be a facility by which a sender can be
   assured that an EDI transmission was successfully
   encrypted, signed, and sent.

4.3.2 Recommendations
   * This should be required functionality of the UA.

   * The UA needs to be able to identify the transmission
   by its interchange control #, AND a user defined value,
   if desired.

4.4 Section: Transmission successfully received
4.4.1 Need
   There needs to be a facility by which a sender of a
   transmission can be assured that the transmission was
   correctly received by the intended receiver.

4.4.2  Recommendations
   1) The UA should track this, and provide the sender with
   signed receipts.

   2) The use of the MDN (message disposition notification)
   as described in Section 3.7.3, according to the Internet
   draft by Roger Fajman

   3) Note:  This could theoretically be accomplished by
   using a 997 in place of the NRR, however, it's our
   recommendation to not do that for two reasons:

          * The implied success of the receiver's decryption
          through the receipt of a legible 997, binds the
          certificate to a control ID only (997) and not to
          the actual data (NRR).

          * Translators are very different, and we feel this
          RFC should define interoperability between UA's
          only, and still cover all angles.

4.5 Section: Transmission successfully translated by
4.5.1 Need
   There needs to be a facility for the sender to be
   assured that the receiver could "understand" (in EDI
   terms) the transmission.

4.5.2 Recommendations
   This should NOT be tracked by the UA, following our
   recommendation for object boundaries

   The Functional acknowledgment 997, and the EDIFACT
   CONTRL serve this exact purpose - this should be tracked
   by the EDI translator.

4.6  Detection and recovery of delayed or lost transmissions
4.6.1  Need
   There needs to be a facility by which a sender can
   detect sent transmissions that have not been
   acknowledged as correctly received, by a specified,
   configurable, period of time, and be able to configure
   actions accordingly.

4.6.2  Recommendations
   1) The use of time stamps for each of the two events:

          * MIME message sent.

          * Signed Receipt received.

   2) The ability to automatically detect transmissions
   that have failed the time trigger.

   3) The ability to configure automatic actions based on
   failure.  Actions may include:

          * Re-transmit.  If re-transmitted, the receiving
          UA needs to be able to detect the second
          transmission as a duplicate and discard it (more
          on this below).
                    * Alert/Report.
          * Ignore/delete (this option may be chosen by
          someone that has decided to track only at the EDI
          translator level through 997/CONTRL.

4.7  Detection and handling of duplicate transmissions
4.7.1  Need
   There needs to be a facility by which a receiver of EDI
   transmissions is able to detect different types of
   duplicate transmissions, and handle them the way they
   should be handled.  First, translator initiated
   duplicates should NOT be halted in any way - we should
   assume that translators will handle that level.  In
   other words, there should be no checking of ISA control
   numbers by the UA.  Secondly, the use of a re-
   transmission feature in attempts to deliver
   transmissions quickly, should allow for a UA to identify
   duplicate transmissions generated by the sending UA, and
   discard of duplicate transmissions after the first has
   been received.

4.7.1  Need
   The ability to pass through translator initiated re-
   transmissions to the receiving translator without a
   hitch.  This means EDI related control numbers, such as
   the ISA control number, should not be checked by the UA.

Appendix A - A Comparison of Security Protocols

                         Version: 3.0

                     Date: July 18, 1996


EDIINT- EDI over Internet, Internet Mail Consortium Workshop
data, Chuck Shih, Steve Dusse', David Darnell, Kent
Landfield, David Chia, Rik Drummond, Jeff Cook, Alan Cox,
Raph Levien, Russ Housley, and many others.

1) Exportable Out Side Of The USA
PGP V3.0
    * PGP is already outside the USA and except for
     countries that prohibit encrypted messages with long
     key lengths (instead of just restricting the import of
     long key length algorithms) PGP long key lengths
     messages can be read This is included in the PGP
     ViaCrypt documentation:
    * No since the encryption algorithm specified is IDEA.

    * Has the 40 and 56 bit export restrictions if RC2 or
     RC4 is used for encryption

    * Not with full key length
    * Depends on the data encryption algorithm used. RFC
     1423 specifies DES in CBC mode, which is not
     exportable. Moss however allows the use of variety of
     cryptographic algorithms.

    * Depends on the  key management and data encryption
     algorithm used. MSP allows the use of variety of
     cryptographic algorithms.

2) Easily Integrated Into Products In A User Transparent
PGP V3.0
    * Maybe in V.30. Not in earlier versions
    * There seems to be general disagreement on this one

    * Yes

    * Do  not know

    * Yes.  Support for signed receipts may require GUI

3) Fully Compatible With Like Versions World Wide
PGP V3.0
    * PGP version 2.6 is compatible with any earlier
     versions. Version 3.0 should be also.

    * RSA has an active interoperability program in place
    * Implementations to the spec should guarantee

    * Moss does not require any particular security
     algorithm. Moss provides the means to identify which
     algorithms are used for each message. A suite of
     algorithms is defined in RFC 1423.

    * Implementations to the spec should guarantee
     interoperability when the same cryptography is used.

4) Current Implementation Status
PGP V3.0
    * Version 3.0 is out 3Q96
    * Version 2.6 is available
    * Qualcomm
    * Premail
    * Michael's PGPMIME

    * Two companies have  implemented several others have
    * Product is shipping

    * TIS, Innosoft and SupplyTech

    * SPYRUS, Nortel, Xerox, LJL, BBN, and J. G. Van Dyke
     all have implementations.
    * Product is shipping.
    * In use for Military messages.

5) Confidentiality
PGP V3.0
    * YesS/MIME    * YesMOSS
    * Yes

    * Yes

6) Signature
PGP V3.0
    * Yes

    * Yes

    * Yes

    * Yes

7) Return Receipt
PGP V3.0
    * Via MIME extensions RFC1891-94

    * Via MIME extensions RFC1891-94

    * Via MIME extensions RFC1891-94

    * Yes.; supports non-repudiation with proof of delivery.

8) Delivery Notification
PGP V3.0
    * Via MIME extensions RFC1891-94

    * Via MIME extensions RFC1891-94

    * Via MIME extensions RFC1891-94

    * Via MIME extensions RFC1891-94

9) Authentication
PGP V3.0
    * Yes

    * Yes

    * Yes

    * Yes

10) Multimedia
PGP V3.0
    * Yes


    * Yes

    * Yes

11) Integrity
PGP V3.0
    * Yes

    * Yes

    * Yes

    * Yes

12) Trust Model (Key  Management & Revocation)
PGP V3.0    * PGP 3.0 will have hierarchical model of public-key
   * RSA used for key management in current versions
    * Ad hoc key revocation.

    * RSA based using X.509 all versions.
    * NT's Entrust will be usable with this product very

    * Both RSA and DES based key management.

    * X.509 all versions.

13) Certificate (Information, Format, Distribution)
PGP V3.0
    * Yes using proprietary "Key rings". Not clear what V3
     will use

    * Yes using X.509 -all versions

    * Yes with optional X.509

    * Yes using X.509

14) Infrastructure Overhead
PGP V3.0
    * Base 64 encoding

    * ASN.1 - BER and DER encoding
    * Base 64 encoding

    * Base 64 encoding

    * Base64 encoding and  ASN.1 encoding

15) Envelope Type
PGP V3.0

    * PKCS #7 ASN.1 and MIME


    * MIME /ASN.1

16) Envelope / Structure Components (ASN1 Or ASCII)
PGP V3.0
    * ASCII

    * ASN.1and ASCII

    * ASCII

    * ASN.1

17) Algorithms Supported (List Them: Encryption,  Key
Management,  One Way Hash,  Digital Signature, Key Lengths
     For Encryption)
PGP V3.0
    * RSA and IDEA in pre 3.0
    * Diffie Hellman and DSA in Ver. 3.0
    * I DEA in CBC
    * MD5 & RSA
    * A  384 for casual grade, 512 commercial grade, 2048
     military grade
    * A 128 bit IDEA key length

    * RSA
    * RC2 / RC5
    * MD5 & RSA
    * SHA-V
   * Note: S/MIME like Moss is a format and allows any type
     of algorithm to be specified. RSA of course specifies
     their own algorithms
    * Triple-DES/RC5

    * DES in CBC
    * RSA or DES
    * MD2/MD5 and RSA
    * A 56 bit key lengths for DES
    * Note: Moss like S/MIME allows a variety of
     cryptographic algorithms to be used. The suite of
     algorithms defined above are found in RFC 1423.

    * Algorithm independent.  Implementations exist  using:
    * RSA & DES
    * FORTEZZA (DSS SHA-1, KEA, Skipjack)

18) Common Algorithms With EDIFACT AUTACK List Of Codes
PGP V3.0
    * RSA  (yes)
    * IDEA (yes)
    * DH (?)
    * DSA (yes)
    * MD5 (yes)

    * RSA (yes)
    * RC2 and RC4 (yes)
    * DES (yes)
    * MD5 (yes)

    * RSA (yes)
    * DES (yes)
    * MD5 (yes)

    * RSA (yes)
    * DES (yes)
    * MD5 (yes)
    * DSS (yes)
    * SHA-1 (yes)

19) Coexistence With Others For Reception (signature not
     readable) Of MIME Multipart/Signed Data
PGP V3.0
    * Yes V3.0

    * Yes, but user selectable

    * Yes

    * Yes, if used with MIME encapsulation (see  <draft-

20) Signed Message  Body Readable By RFC822/ MIME Readers
PGP V3.0
    * Should be in V3.0

    * Yes -  if one of the options multipart/signed is used
    * Verify with RSA is it multipart/alternative and not

    * Yes

    * Yes, if used with MIME encapsulation

21) Signature Separate From Signed Document
PGP V3.0
    * Yes

    * Yes, optional
    * Verify with RSA

    * Yes

    * Yes

22) Backward Compatibility
PGP V3.0
    * To PGP

    * To PEM

    * To PEM

    * No

23) Uses Proprietary Algorithms?
PGP V3.0
    * Ver 3.0 will use Diffie-Hellman with expiring patents
     in 1997.  Ver 3.0 will use DSA (Digital Signature
     Algorithm invented at the NSA)

    * Yes

    * YES, but it supports different options in a coherent

    * It supports both standard (FIPS and X9) as well as
     proprietary algorithms.

24) Adequate Security For EDI Purposes
PGP V3.0
    * Yes

    * Yes. However one can tell the difference between an
     encrypted message and a signed/encrypted message.
     There is not consensus as to if this is a problem or

    * Yes

    * Yes

25) Scaleable
PGP V3.0
    * No enough experience to tell. The current trust model
     will not scale well.

    * No enough experience to tell

    * No enough experience to tell

    * Yes

26) Solid Mime Integration
PGP V3.0
    * V3.0 -yes

    * Yes - but it mixes PKCS technology with MIME. Internet
     purest do not seem to like the mix.

    * Yes

    * Yes  (see  <draft-housley-msp-mime-01.txt>

27) Variable Key Sizes Supported
PGP V3.0

    * Yes, 40- 128 bit
    * Symmetric 512-2048 bit RSA


    * Yes
    * Symmetric (DES = 56 bits; SKIPJACK = 80 bits)
    * Signature (DSS = 512 .. 1024 bits; RSA = 512 .. 2048
    * Key Management (KEA = 1024 bits; RSA = 512 .. 2048

28) Only X.509 Or Other Certificate Distribution Methods
PGP V3.0

    * X.509 any version


    * X.509 any version

29) Very Solid API And Took Kit
PGP V3.0
    * V3.0 Yes - anticipated

    * Yes

    * No

    * Yes.  DISA is submitting an API to X/Open.
    * At least two tookkits for sale.

30) Tool Kits Can Be Made Available To All Countries Not
     Under UN Embargo
PGP V3.0
    * Probably from alternate sources

    * Export version probably to most countries

    * Export version probably to most countries (40 bit
    * Alternate sources probably yes

    * Export version probably to most countries.  Toolkits
     exported to many countries, including France.
    * Alternate source likely.

31) Fit With Future Direction Of EDI
PGP V3.0
    * Neutral

    * Neutral

    * Neutral

    * Neutral


Chuck Shih <cshih@netscape.com>
  Phone: 415  937 3511 USA


Rik Drummond <drummond@onramp.net>
  Phone: 817 294 7339   USA