OAuth Working Group N. Sakimura, Ed.
Internet-Draft Nomura Research Institute
Intended status: Standards Track J. Bradley
Expires: December 8, 2015 Ping Identity
N. Agarwal
June 6, 2015

Proof Key for Code Exchange by OAuth Public Clients


OAuth 2.0 public clients utilizing the Authorization Code Grant are susceptible to the authorization code interception attack. This specification describes the attack as well as a technique to mitigate against the threat.

Status of This Memo

This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.

Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at http://datatracker.ietf.org/drafts/current/.

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."

This Internet-Draft will expire on December 8, 2015.

Copyright Notice

Copyright (c) 2015 IETF Trust and the persons identified as the document authors. All rights reserved.

This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.

Table of Contents

1. Introduction

OAuth 2.0 [RFC6749] public clients are susceptible to the authorization code interception attack.

The attacker thereby intercepts the authorization code returned from the authorization endpoint within communication path not protected by TLS, such as inter-app communication within the operating system of the client.

Once the attacker has gained access to the authorization code it can use it to obtain the access token.

Figure 1 shows the attack graphically. In step (1) the native app running on the end device, such as a smart phone, issues an authorization request via the browser/operating system, which then gets forwarded to the OAuth 2.0 authorization server in step (2). The authorization server returns the authorization code in step (3). The malicious app is able to observe the authorization code in step (4) since it is registered to the custom URI scheme used by the legitimate app. This allows the attacker to reguest and obtain an access token in step (5) and step (6), respectively.

 | End Device (e.g., Smart Phone) |
 |                                |
 | +-------------+   +----------+ | (6) Access Token  +----------+
 | |Legitimate   |   | Malicious|<--------------------|          |
 | |OAuth 2.0 App|   | App      |-------------------->|          |
 | +-------------+   +----------+ | (5) Authorization |          |
 |        |    ^          ^       |        Grant      |          |
 |        |     \         |       |                   |          |
 |        |      \   (4)  |       |                   |          |
 |    (1) |       \  Authz|       |                   |          |
 |   Authz|        \ Code |       |                   |  Authz   |
 | Request|         \     |       |                   |  Server  |
 |        |          \    |       |                   |          |
 |        |           \   |       |                   |          |
 |        v            \  |       |                   |          |
 | +----------------------------+ |                   |          |
 | |                            | | (3) Authz Code    |          |
 | |     Operating System/      |<--------------------|          |
 | |         Browser            |-------------------->|          |
 | |                            | | (2) Authz Request |          |
 | +----------------------------+ |                   +----------+

Figure 1: Authorization Code Interception Attack.

A number of pre-conditions need to hold in order for this attack to work:

The attacker manages to register a malicious application on the client device and registers a custom URI scheme that is also used by another application.
The operating systems must allow a custom URI schemes to be registered by multiple applications.
The OAuth 2.0 authorization code grant is used.
The attacker has access to the OAuth 2.0 [RFC6749] client_id and client_secret(if provisioned). All OAuth 2.0 native app client-instances use the same client_id. Secrets provisioned in client binary applications cannot be considered confidential.
The attacker (via the installed app) is able to observe responses from the authorization endpoint. As a more sophisticated attack scenario the attacker is also able to observe requests (in addition to responses) to the authorization endpoint. The attacker is, however, not able to act as a man-in-the-middle.

While this is a long list of pre-conditions the described attack has been observed in the wild and has to be considered in OAuth 2.0 deployments.
While the OAuth 2.0 Threat Model Section 4.4.1 [RFC6819] describes mitigation techniques they are, unfortunately, not applicable since they rely on a per-client instance secret or aper client instance redirect URI.

To mitigate this attack, this extension utilizes a dynamically created cryptographically random key called 'code verifier'. A unique code verifier is created for every authorization request and its transformed value, called 'code challenge', is sent to the authorization server to obtain the authorization code. The authorization code obtained is then sent to the token endpoint with the 'code verifier' and the server compares it with the previously received request code so that it can perform the proof of possession of the 'code verifier' by the client. This works as the mitigation since the attacker would not know this one-time key.

1.1. Protocol Flow

                                              |   Authz Server    |
    +--------+                                | +---------------+ |
    |        |--(A)- Authorization Request ---->|               | |
    |        |        + t(code_verifier), t   | | Authorization | |
    |        |                                | |    Endpoint   | |
    |        |<-(B)---- Authorization Code -----|               | |
    |        |                                | +---------------+ |
    | Client |                                |                   |
    |        |                                | +---------------+ |
    |        |--(C)-- Access Token Request ---->|               | |
    |        |          + code_verifier       | |    Token      | |
    |        |                                | |   Endpoint    | |
    |        |<-(D)------ Access Token ---------|               | |
    +--------+                                | +---------------+ |

Figure 2: Abstract Protocol Flow

This specification adds additional parameters to the OAuth 2.0 Authorization and Access Token Requests, shown in abstract form in Figure 1.

The client creates and records a secret named the code_verifier, and derives a transformed version t(code_verifier) (referred to as the code_challenge) which is sent in the OAuth 2.0 Authorization Request, along with the transformation method t.
The Authorization Endpoint responds as usual, but records t(code_verifier) and the transformation method.
The client then sends the code in the Access Token Request as usual, but includes the code_verifier secret generated at (A).
The authorization server transforms code_verifier and compares it to t(code_verifier) from (B). Access is denied if they are not equal.

An attacker who intercepts the Authorization Grant at (B) is unable to redeem it for an Access Token, as they are not in possession of the code_verifier secret.

2. Notational Conventions

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in Key words for use in RFCs to Indicate Requirement Levels [RFC2119]. If these words are used without being spelled in uppercase then they are to be interpreted with their normal natural language meanings.

This specification uses the Augmented Backus-Naur Form (ABNF) notation of [RFC5234].

STRING denotes a sequence of zero or more ASCII [RFC0020] characters.

OCTETS denotes a sequence of zero or more octets.

ASCII(STRING) denotes the octets of the ASCII [RFC0020] representation of STRING where STRING is a sequence of zero or more ASCII characters.

BASE64URL-ENCODE(OCTETS) denotes the base64url encoding of OCTETS, per Section 3 producing a STRING.

BASE64URL-DECODE(STRING) denotes the base64url decoding of STRING, per Section 3, producing a sequence of octets.

SHA256(OCTETS) denotes a SHA2 256bit hash [RFC6234] of OCTETS.

3. Terminology

In addition to the terms defined in OAuth 2.0 [RFC6749], this specification defines the following terms:

code verifier
A cryptographically random string that is used to correlate the authorization request to the token request.
code challenge
A challenge derived from the code verifier that is sent in the authorization request, to be verified against later.
Base64url Encoding
Base64 encoding using the URL- and filename-safe character set defined in Section 5 of [RFC4648], with all trailing '=' characters omitted (as permitted by Section 3.2 of [RFC4648]) and without the inclusion of any line breaks, whitespace, or other additional characters. (See Appendix A for notes on implementing base64url encoding without padding.)

4. Protocol

4.1. Client creates a code verifier

The client first creates a code verifier, code_verifier, for each OAuth 2.0 [RFC6749] Authorization Request, in the following manner:

code_verifier = high entropy cryptographic random STRING using the Unreserved Characters [A-Z] / [a-z] / [0-9] / "-" / "." / "_" / "~" from Sec 2.3 of [RFC3986], with a minimum length of 43 characters and a maximum length of 128 characters.

ABNF for code_verifier is as follows.

code-verifier = 43*128unreserved
unreserved = ALPHA / DIGIT / "-" / "." / "_" / "~"
ALPHA = %x41-5A / %x61-7A
DIGIT = %x30-39

NOTE: code verifier SHOULD have enough entropy to make it impractical to guess the value. It is RECOMMENDED that the output of a suitable random number generator be used to create a 32-octet sequence. The Octet sequence is then base64url encoded to produce a 43-octet URL safe string to use as the code verifier.

4.2. Client creates the code challenge

The client then creates a code challenge, code_challenge, derived from the code_verifier by using one of the following transformations on the code_verifier:

code_challenge = code_verifier
code_challenge = BASE64URL-ENCODE(SHA256(ASCII(code_verifier)))

It is RECOMMENDED to use the S256 transformation when possible.

ABNF for code_challenge is as follows.

code-challenge = 43*128unreserved
unreserved = ALPHA / DIGIT / "-" / "." / "_" / "~"
ALPHA = %x41-5A / %x61-7A
DIGIT = %x30-39

4.3. Client sends the code challenge with the authorization request

The client sends the code challenge as part of the OAuth 2.0 Authorization Request (Section 4.1.1 of [RFC6749].) using the following additional parameters:

REQUIRED. Code challenge.

OPTIONAL, defaults to plain. Code verifier transformation method, S256 or plain.

4.4. Server returns the code

When the server issues the code in the Authorization Response, it MUST associate the code_challenge and code_challenge_method values with the code so it can be verified later.

Typically, the code_challenge and code_challenge_method values are stored in encrypted form in the code itself, but could alternatively be stored on the server, associated with the code. The server MUST NOT include the code_challenge value in client requests in a form that other entities can extract.

The exact method that the server uses to associate the code_challenge with the issued code is out of scope for this specification.

4.4.1. Error Response

If the server requires PKCE, and the client does not send the code_challenge in the request, the authorization endpoint MUST return the authorization error response with error value set to invalid_request. The error_description or the response of error_uri SHOULD explain the nature of error, e.g., code challenge required.

If the server supporting PKCE does not support the requested transform, the authorization endpoint MUST return the authorization error response with error value set to invalid_request. The error_description or the response of error_uri SHOULD explain the nature of error, e.g., transform algorithm not supported.

If the client is capable of using S256, it MUST use S256, as S256 is Mandatory To Implement (MTI) on the server. Clients MAY use plain only if they cannot support S256 for some technical reason and knows that the server supports plain.

4.5. Client sends the code and the secret to the token endpoint

Upon receipt of the code, the client sends the Access Token Request to the token endpoint. In addition to the parameters defined in the OAuth 2.0 Access Token Request (Section 4.1.3 of [RFC6749]), it sends the following parameter:

REQUIRED. Code verifier

4.6. Server verifies code_verifier before returning the tokens

Upon receipt of the request at the Access Token endpoint, the server verifies it by calculating the code challenge from received code_verifier and comparing it with the previously associated code_challenge, after first transforming it according to the code_challenge_method method specified by the client.

If the code_challenge_method from Section 4.2 was S256, the received code_verifier is hashed by SHA-256, then base64url encoded, and then compared to the code_challenge. i.e.,

BASE64URL-ENCODE(SHA256(ASCII(code_verifier ))) == code_challenge

If the code_challenge_method from Section 4.2 was plain, they are compared directly. i.e.,

code_verifier == code_challenge.

If the values are equal, the Access Token endpoint MUST continue processing as normal (as defined by OAuth 2.0 [RFC6749]). If the values are not equal, an error response indicating invalid_grant as described in section 5.2 of [RFC6749] MUST be returned.

5. Compatibility

Server implementations of this specification MAY accept OAuth2.0 Clients that do not implement this extension. If the code_verifier is not received from the client in the Authorization Request, servers supporting backwards compatibility SHOULD revert to a normal OAuth 2.0 [RFC6749] protocol.

As the OAuth 2.0 [RFC6749] server responses are unchanged by this specification, client implementations of this specification do not need to know if the server has implemented this specification or not, and SHOULD send the additional parameters as defined in Section 3. to all servers.

6. IANA Considerations

This specification makes a registration request as follows:

6.1. OAuth Parameters Registry

This specification registers the following parameters in the IANA OAuth Parameters registry defined in OAuth 2.0 [RFC6749].

  • Parameter name: code_verifier
  • Parameter usage location: Access Token Request
  • Change controller: IESG
  • Specification document(s): this document

  • Parameter name: code_challenge
  • Parameter usage location: Authorization Request
  • Change controller: IESG
  • Specification document(s): this document

  • Parameter name: code_challenge_method
  • Parameter usage location: Authorization Request
  • Change controller: IESG
  • Specification document(s): this document

6.2. PKCE Code Challenge Method Registry

This specification establishes the PKCE Code Challenge Method registry.

Additional code_challenge_method types for use with the authorization endpoint are registered with a Specification Required ([RFC5226]) after a two-week review period on the oauth-ext-review@ietf.org mailing list, on the advice of one or more Designated Experts. However, to allow for the allocation of values prior to publication, the Designated Expert(s) may approve registration once they are satisfied that such a specification will be published.

Registration requests must be sent to the oauth-ext-review@ietf.org mailing list for review and comment, with an appropriate subject (e.g., "Request for PKCE code_challenge_method: example").

Within the review period, the Designated Expert(s) will either approve or deny the registration request, communicating this decision to the review list and IANA. Denials should include an explanation and, if applicable, suggestions as to how to make the request successful.

IANA must only accept registry updates from the Designated Expert(s) and should direct all requests for registration to the review mailing list.

6.2.1. Registration Template

Code Challenge Method Parameter Name:

The name requested (e.g., "example"). Because a core goal of this specification is for the resulting representations to be compact, it is RECOMMENDED that the name be short -- not to exceed 8 characters without a compelling reason to do so. This name is case-sensitive. Names may not match other registered names in a case-insensitive manner unless the Designated Expert(s) state that there is a compelling reason to allow an exception in this particular case.
Change Controller:

For Standards Track RFCs, state "IESG". For others, give the name of the responsible party. Other details (e.g., postal address, email address, home page URI) may also be included.
Specification Document(s):

Reference to the document(s) that specify the parameter, preferably including URI(s) that can be used to retrieve copies of the document(s). An indication of the relevant sections may also be included but is not required.

6.2.2. Initial Registry Contents

This specification registers the Code Challenge Method Parameter names defined in Section 4.2 in this registry.

  • Code Challenge Method Parameter Name: plain
  • Change Controller: IESG
  • Specification Document(s): Section 4.2 of [[ this document ]]

  • Code Challenge Method Parameter Name: S256
  • Change Controller: IESG
  • Specification Document(s): Section 4.2 of [[ this document ]]

7. Security Considerations

7.1. Entropy of the code_verifier

The security model relies on the fact that the code verifier is not learned or guessed by the attacker. It is vitally important to adhere to this principle. As such, the code verifier has to be created in such a manner that it is cryptographically random and has high entropy that it is not practical for the attacker to guess.

The client SHOULD create a code_verifier with a minimum of 256bits of entropy. This can be done by having a suitable random number generator create a 32-octet sequence. The Octet sequence can then be base64url encoded to produce a 43-octet URL safe string to use as a code_challenge that has the required entropy.

7.2. Protection against eavesdroppers

Clients MUST NOT try down grading the algorithm after trying S256 method. If the server is PKCE compliant, then S256 method works. If the server does not support PKCE, it does not generate error. Only the time that the server returns that it does not support S256 is there is a MITM trying the algorithm downgrade attack.

S256 method protects against eavesdroppers observing or intercepting the code_challenge. If the plain method is used, there is a chance that it will be observed by the attacker on the device. The use of S256 protects against it.

If code_challenge is to be returned inside authorization code to achieve a stateless server, it has to be encrypted in such a manner that only the server can decrypt and extract it.

7.3. Salting the code_challenge

In order to reduce implementation complexity Salting is not used in the production of the code_challenge, as the code_verifier contains sufficient entropy to prevent brute force attacks. Concatenating a publicly known value to a code_verifier (containing 256 bits of entropy) and then hashing it with SHA256 to produce a code_challenge would not increase the number of attempts necessary to brute force a valid value for code_verifier.

While the S256 transformation is like hashing a password there are important differences. Passwords tend to be relatively low entropy words that can be hashed offline and the hash looked up in a dictionary. By concatenating a unique though public value to each password prior to hashing, the dictionary space that an attacker needs to search is greatly expanded.

Modern graphics processors now allow attackers to calculate hashes in real time faster than they could be looked up from a disk. This eliminates the value of the salt in increasing the complexity of a brute force attack for even low entropy passwords.

7.4. OAuth security considerations

All the OAuth security analysis presented in [RFC6819] applies so readers SHOULD carefully follow it.

7.5. TLS security considerations

Curent security considerations can be found in Recommendations for Secure Use of TLS and DTLS [BCP195]. This supersedes the TLS version recommendations in OAuth 2.0 [RFC6749].

8. Acknowledgements

The initial draft of this specification was created by the OpenID AB/Connect Working Group of the OpenID Foundation.

This specification is the work of the OAuth Working Group, which includes dozens of active and dedicated participants. In particular, the following individuals contributed ideas, feedback, and wording that shaped and formed the final specification:

  • Anthony Nadalin, Microsoft
  • Axel Nenker, Deutsche Telekom
  • Breno de Medeiros, Google
  • Brian Campbell, Ping Identity
  • Chuck Mortimore, Salesforce
  • Dirk Balfanz, Google
  • Eduardo Gueiros, Jive Communications
  • Hannes Tschonfenig, ARM
  • James Manger, Telstra
  • John Bradley, Ping Identity
  • Justin Richer, MIT Kerberos
  • Josh Mandel, Boston Children's Hospital
  • Lewis Adam, Motorola Solutions
  • Madjid Nakhjiri, Samsung
  • Michael B. Jones, Microsoft
  • Nat Sakimura, Nomura Research Institute
  • Naveen Agarwal, Google
  • Paul Madsen, Ping Identity
  • Phil Hunt, Oracle
  • Prateek Mishra, Oracle
  • Ryo Ito, mixi
  • Scott Tomilson, Ping Identity
  • Sergey Beryozkin
  • Takamichi Saito
  • Torsten Lodderstedt, Deutsche Telekom
  • William Denniss, Google

9. Revision History


  • clarify that the client secret we are talking about in the Introduction is a OAuth 2 client_secret.
  • Update salting security consideration based on Ben's feedback


  • add spanx for plain in sec 4.4 RE Kathleen's comment
  • Add security consideration on TLS and reference BCP195


  • re #33 specify lower limit to code_verifier in prose
  • remove base64url decode from draft, all steps now use encode only
  • Expanded MTI
  • re #33 change length of 32 octet base64url encoded string back to 43 octets


  • clean up some external references so they don't point at internal sections


  • changed BASE64URL to BASE64URL-ENCODE to be more consistent with appendix A Fixed lowercase base64url in appendix B
  • Added appendix B as an example of S256 processing
  • Change reference for unreserved characters to RFC3986 from base64URL


  • removed unused discovery reference and UTF8
  • re #32 added ASCII(STRING) to make clear that it is the byte array that is being hashed
  • re #2 Remove discovery requirement section.
  • updated Acknowledgement
  • re #32 remove unneeded UTF8(STRING) definition, and define STRING for ASCII(STRING)
  • re #32 remove unneeded utf8 reference from BASE64URL-DECODE(STRING) def
  • resolves #31 unused definition of concatenation
  • re #30 Update figure text call out the endpoints
  • re #30 Update figure to call out the endpoints
  • small wording change to the introduction


  • fix date
  • replace spop with pkce for registry and other references
  • re #29 change name again
  • re #27 removed US-ASCII reference
  • re #27 updated ABNF for code_verifier
  • resolves #24 added security consideration for salting
  • resolves #29 Changed title
  • updated reference to RFC4634 to RFC6234 re #27
  • changed reference for US-ASCII to RFC20 re #27
  • resolves #28 added Acknowledgements
  • resolves #27 updated ABNF
  • resolves #26 updated abstract and added Hannes figure


  • Added IANA registry for code_challenge_method + fixed some broken internal references.


  • Added error response to authorization response.


  • Added an abstract protocol diagram and explanation


  • Copy edits


  • Specified exactly two supported transformations
  • Moved discovery steps to security considerations.
  • Incorporated readability comments by Eduardo Gueiros.
  • Changed MUST in 3.1 to SHOULD.


  • Initial IETF version.

10. References

10.1. Normative References

[BCP195] Sheffer, Y., Holz, R. and P. Saint-Andre, "Recommendations for Secure Use of Transport Layer Security (TLS) and Datagram Transport Layer Security (DTLS)", BCP 195, RFC 7525, May 2015.
[RFC0020] Cerf, V., "ASCII format for network interchange", RFC 20, October 1969.
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997.
[RFC3986] Berners-Lee, T., Fielding, R. and L. Masinter, "Uniform Resource Identifier (URI): Generic Syntax", STD 66, RFC 3986, January 2005.
[RFC4648] Josefsson, S., "The Base16, Base32, and Base64 Data Encodings", RFC 4648, October 2006.
[RFC5234] Crocker, D. and P. Overell, "Augmented BNF for Syntax Specifications: ABNF", STD 68, RFC 5234, January 2008.
[RFC6234] Eastlake, D. and T. Hansen, "US Secure Hash Algorithms (SHA and SHA-based HMAC and HKDF)", RFC 6234, May 2011.
[RFC6749] Hardt, D., "The OAuth 2.0 Authorization Framework", RFC 6749, October 2012.

10.2. Informative References

[RFC6819] Lodderstedt, T., McGloin, M. and P. Hunt, "OAuth 2.0 Threat Model and Security Considerations", RFC 6819, January 2013.

Appendix A. Notes on implementing base64url encoding without padding

This appendix describes how to implement a base64url encoding function without padding based upon standard base64 encoding function that uses padding.

To be concrete, example C# code implementing these functions is shown below. Similar code could be used in other languages.

  static string base64urlencode(byte [] arg)
    string s = Convert.ToBase64String(arg); // Regular base64 encoder
    s = s.Split('=')[0]; // Remove any trailing '='s
    s = s.Replace('+', '-'); // 62nd char of encoding
    s = s.Replace('/', '_'); // 63rd char of encoding
    return s;

An example correspondence between unencoded and encoded values follows. The octet sequence below encodes into the string below, which when decoded, reproduces the octet sequence.

3 236 255 224 193

Appendix B. Example for the S256 code_challenge_method

The client uses output of a suitable random number generator to create a 32-octet sequence. The octets representing the value in this example (using JSON array notation) are:"

   [116, 24, 223, 180, 151, 153, 224, 37, 79, 250, 96, 125, 216, 173,
   187, 186, 22, 212, 37, 77, 105, 214, 191, 240, 91, 88, 5, 88, 83,
   132, 141, 121]

Encoding this octet sequence as a Base64url provides the value of the code_verifier:


The code_verifier is then hashed via the SHA256 hash function to produce:

  [19, 211, 30, 150, 26, 26, 216, 236, 47, 22, 177, 12, 76, 152, 46,
   8, 118, 168, 120, 173, 109, 241, 68, 86, 110, 225, 137, 74, 203, 
   112, 249, 195]

Encoding this octet sequence as a base64url provides the value of the code_challenge:


The authorization request includes:


The Authorization server then records the code_challenge and code_challenge_method along with the code that is granted to the client.

in the request to the token_endpoint the client includes the code received in the authorization response as well as the additional paramater:


The Authorization server retrieves the information for the code grant. Based on the recorded code_challange_method being S256, it then hashes and base64url encodes the value of code_verifier. BASE64URL-ENCODE(SHA256(ASCII(code_verifier )))

The calculated value is then compared with the value of code_challenge:

BASE64URL-ENCODE(SHA256(ASCII(code_verifier ))) == code_challenge

If the two values are equal then the Authorization server can provide the tokens as long as there are no other errors in the request. If the values are not equal then the request must be rejected, and an error returned.

Authors' Addresses

Nat Sakimura (editor) Nomura Research Institute 1-6-5 Marunouchi, Marunouchi Kitaguchi Bldg. Chiyoda-ku, Tokyo 100-0005 Japan Phone: +81-3-5533-2111 EMail: n-sakimura@nri.co.jp URI: http://nat.sakimura.org/
John Bradley Ping Identity Casilla 177, Sucursal Talagante Talagante, RM Chile Phone: +44 20 8133 3718 EMail: ve7jtb@ve7jtb.com URI: http://www.thread-safe.com/
Naveen Agarwal Google 1600 Amphitheatre Pkwy Mountain View, CA 94043 USA Phone: +1 650-253-0000 EMail: naa@google.com URI: http://google.com/