securityos/node_modules/tough-cookie

tough-cookie

RFC 6265 Cookies and CookieJar for Node.js

Synopsis

var tough = require("tough-cookie");
var Cookie = tough.Cookie;
var cookie = Cookie.parse(header);
cookie.value = "somethingdifferent";
header = cookie.toString();
var cookiejar = new tough.CookieJar();

// Asynchronous!
var cookie = await cookiejar.setCookie(
  cookie,
  "https://currentdomain.example.com/path"
);
var cookies = await cookiejar.getCookies("https://example.com/otherpath");

// Or with callbacks!
cookiejar.setCookie(
  cookie,
  "https://currentdomain.example.com/path",
  function (err, cookie) {
    /* ... */
  }
);
cookiejar.getCookies("http://example.com/otherpath", function (err, cookies) {
  /* ... */
});

Why the name? NPM modules cookie, cookies and cookiejar were already taken.

Installation

It's so easy! Install with npm or your preferred package manager.

npm install tough-cookie

Node.js Version Support

We follow the node.js release schedule and support all versions that are in Active LTS or Maintenance. We will always do a major release when dropping support for older versions of node, and we will do so in consultation with our community.

API

tough

The top-level exports from require('tough-cookie') can all be used as pure functions and don't need to be bound.

parseDate(string)

Parse a cookie date string into a Date. Parses according to RFC 6265 Section 5.1.1, not Date.parse().

formatDate(date)

Format a Date into an RFC 822 string (the RFC 6265 recommended format).

canonicalDomain(str)

Transforms a domain name into a canonical domain name. The canonical domain name is a domain name that has been trimmed, lowercased, stripped of leading dot, and optionally punycode-encoded (Section 5.1.2 of RFC 6265). For the most part, this function is idempotent (calling the function with the output from a previous call returns the same output).

domainMatch(str, domStr[, canonicalize=true])

Answers "does this real domain match the domain in a cookie?". The str is the "current" domain name and the domStr is the "cookie" domain name. Matches according to RFC 6265 Section 5.1.3, but it helps to think of it as a "suffix match".

The canonicalize parameter toggles whether the domain parameters get normalized with canonicalDomain or not.

defaultPath(path)

Given a current request/response path, gives the path appropriate for storing in a cookie. This is basically the "directory" of a "file" in the path, but is specified by Section 5.1.4 of the RFC.

The path parameter MUST be only the pathname part of a URI (excluding the hostname, query, fragment, and so on). This is the .pathname property of node's uri.parse() output.

pathMatch(reqPath, cookiePath)

Answers "does the request-path path-match a given cookie-path?" as per RFC 6265 Section 5.1.4. Returns a boolean.

This is essentially a prefix-match where cookiePath is a prefix of reqPath.

parse(cookieString[, options])

Alias for Cookie.parse(cookieString[, options]).

fromJSON(string)

Alias for Cookie.fromJSON(string).

getPublicSuffix(hostname)

Returns the public suffix of this hostname. The public suffix is the shortest domain name upon which a cookie can be set. Returns null if the hostname cannot have cookies set for it.

For example: www.example.com and www.subdomain.example.com both have public suffix example.com.

For further information, see the Public Suffix List. This module derives its list from that site. This call is a wrapper around psl's get method.

cookieCompare(a, b)

For use with .sort(), sorts a list of cookies into the recommended order given in step 2 of (RFC 6265 Section 5.4). The sort algorithm is, in order of precedence:

• Longest .path
• oldest .creation (which has a 1-ms precision, same as Date)
• lowest .creationIndex (to get beyond the 1-ms precision)

var cookies = [
  /* unsorted array of Cookie objects */
];
cookies = cookies.sort(cookieCompare);

Note: Since the JavaScript Date is limited to a 1-ms precision, cookies within the same millisecond are entirely possible. This is especially true when using the now option to .setCookie(). The .creationIndex property is a per-process global counter, assigned during construction with new Cookie(), which preserves the spirit of the RFC sorting: older cookies go first. This works great for MemoryCookieStore since Set-Cookie headers are parsed in order, but is not so great for distributed systems. Sophisticated Stores may wish to set this to some other logical clock so that if cookies A and B are created in the same millisecond, but cookie A is created before cookie B, then A.creationIndex < B.creationIndex. If you want to alter the global counter, which you probably shouldn't do, it's stored in Cookie.cookiesCreated.

permuteDomain(domain)

Generates a list of all possible domains that domainMatch() the parameter. Can be handy for implementing cookie stores.

permutePath(path)

Generates a list of all possible paths that pathMatch() the parameter. Can be handy for implementing cookie stores.

Cookie

Exported via tough.Cookie.

Cookie.parse(cookieString[, options])

Parses a single Cookie or Set-Cookie HTTP header into a Cookie object. Returns undefined if the string can't be parsed.

The options parameter is not required and currently has only one property:

• loose - boolean - if true enable parsing of keyless cookies like =abc and =, which are not RFC-compliant.

If options is not an object it is ignored, which means it can be used with Array#map.

To process the Set-Cookie header(s) on a node HTTP/HTTPS response:

if (Array.isArray(res.headers["set-cookie"]))
  cookies = res.headers["set-cookie"].map(Cookie.parse);
else cookies = [Cookie.parse(res.headers["set-cookie"])];

Note: In version 2.3.3, tough-cookie limited the number of spaces before the = to 256 characters. This limitation was removed in version 2.3.4. For more details, see issue #92.

Properties

Cookie object properties:

• key - string - the name or key of the cookie (default "")
• value - string - the value of the cookie (default "")
• expires - Date - if set, the Expires= attribute of the cookie (defaults to the string "Infinity"). See setExpires()
• maxAge - seconds - if set, the Max-Age= attribute in seconds of the cookie. Can also be set to strings "Infinity" and "-Infinity" for non-expiry and immediate-expiry, respectively. See setMaxAge()
• domain - string - the Domain= attribute of the cookie
• path - string - the Path= of the cookie
• secure - boolean - the Secure cookie flag
• httpOnly - boolean - the HttpOnly cookie flag
• sameSite - string - the SameSite cookie attribute (from RFC 6265bis); must be one of none, lax, or strict
• extensions - Array - any unrecognized cookie attributes as strings (even if equal-signs inside)
• creation - Date - when this cookie was constructed
• creationIndex - number - set at construction, used to provide greater sort precision (see cookieCompare(a,b) for a full explanation)

After a cookie has been passed through CookieJar.setCookie() it has the following additional attributes:

• hostOnly - boolean - is this a host-only cookie (that is, no Domain field was set, but was instead implied).
• pathIsDefault - boolean - if true, there was no Path field on the cookie and defaultPath() was used to derive one.
• creation - Date - modified from construction to when the cookie was added to the jar.
• lastAccessed - Date - last time the cookie got accessed. Affects cookie cleaning after it is implemented. Using cookiejar.getCookies(...) updates this attribute.

new Cookie([properties])

Receives an options object that can contain any of the above Cookie properties. Uses the default for unspecified properties.

.toString()

Encodes to a Set-Cookie header value. The Expires cookie field is set using formatDate(), but is omitted entirely if .expires is Infinity.

.cookieString()

Encodes to a Cookie header value (specifically, the .key and .value properties joined with "=").

.setExpires(string)

Sets the expiry based on a date-string passed through parseDate(). If parseDate returns null (that is, can't parse this date string), .expires is set to "Infinity" (a string).

.setMaxAge(number)

Sets the maxAge in seconds. Coerces -Infinity to "-Infinity" and Infinity to "Infinity" so it correctly serializes to JSON.

.expiryDate([now=Date.now()])

expiryTime() computes the absolute unix-epoch milliseconds that this cookie expires. expiryDate() works similarly, except it returns a Date object. Note that in both cases the now parameter should be milliseconds.

Max-Age takes precedence over Expires (as per the RFC). The .creation attribute -- or, by default, the now parameter -- is used to offset the .maxAge attribute.

If Expires (.expires) is set, that's returned.

Otherwise, expiryTime() returns Infinity and expiryDate() returns a Date object for "Tue, 19 Jan 2038 03:14:07 GMT" (latest date that can be expressed by a 32-bit time_t; the common limit for most user-agents).

.TTL([now=Date.now()])

Computes the TTL relative to now (milliseconds). The same precedence rules as for expiryTime/expiryDate apply.

Infinity is returned for cookies without an explicit expiry and 0 is returned if the cookie is expired. Otherwise a time-to-live in milliseconds is returned.

.canonicalizedDomain()

.cdomain()

Returns the canonicalized .domain field. This is lower-cased and punycode (RFC 3490) encoded if the domain has any non-ASCII characters.

.toJSON()

For convenience in using JSON.serialize(cookie). Returns a plain-old Object that can be JSON-serialized.

Any Date properties (such as .expires, .creation, and .lastAccessed) are exported in ISO format (.toISOString()).

NOTE: Custom Cookie properties are discarded. In tough-cookie 1.x, since there was no .toJSON method explicitly defined, all enumerable properties were captured. If you want a property to be serialized, add the property name to the Cookie.serializableProperties Array.

Cookie.fromJSON(strOrObj)

Does the reverse of cookie.toJSON(). If passed a string, will JSON.parse() that first.

Any Date properties (such as .expires, .creation, and .lastAccessed) are parsed via Date.parse, not tough-cookie's parseDate, since ISO timestamps are being handled at this layer.

Returns null upon a JSON parsing error.

.clone()

Does a deep clone of this cookie, implemented exactly as Cookie.fromJSON(cookie.toJSON()).

.validate()

Status: IN PROGRESS. Works for a few things, but is by no means comprehensive.

Validates cookie attributes for semantic correctness. Useful for "lint" checking any Set-Cookie headers you generate. For now, it returns a boolean, but eventually could return a reason string. Future-proof with this construct:

if (cookie.validate() === true) {
  // it's tasty
} else {
  // yuck!
}

CookieJar

Exported via tough.CookieJar.

CookieJar([store][, options])

Simply use new CookieJar(). If a custom store is not passed to the constructor, a MemoryCookieStore is created and used.

The options object can be omitted and can have the following properties:

• rejectPublicSuffixes - boolean - default true - reject cookies with domains like "com" and "co.uk"
• looseMode - boolean - default false - accept malformed cookies like bar and =bar, which have an implied empty name.
• prefixSecurity - string - default silent - set to 'unsafe-disabled', 'silent', or 'strict'. See Cookie Prefixes below.
• allowSpecialUseDomain - boolean - default true - accepts special-use domain suffixes, such as local. Useful for testing purposes. This is not in the standard, but is used sometimes on the web and is accepted by most browsers.

.setCookie(cookieOrString, currentUrl[, options][, callback(err, cookie)])

Attempt to set the cookie in the cookie jar. The cookie has updated .creation, .lastAccessed and .hostOnly properties. And returns a promise if a callback is not provided.

The options object can be omitted and can have the following properties:

• http - boolean - default true - indicates if this is an HTTP or non-HTTP API. Affects HttpOnly cookies.
• secure - boolean - autodetect from URL - indicates if this is a "Secure" API. If the currentUrl starts with https: or wss: this defaults to true, otherwise false.
• now - Date - default new Date() - what to use for the creation or access time of cookies.
• ignoreError - boolean - default false - silently ignore things like parse errors and invalid domains. Store errors aren't ignored by this option.
• sameSiteContext - string - default unset - set to 'none', 'lax', or 'strict' See SameSite Cookies below.

As per the RFC, the .hostOnly property is set if there was no "Domain=" parameter in the cookie string (or .domain was null on the Cookie object). The .domain property is set to the fully-qualified hostname of currentUrl in this case. Matching this cookie requires an exact hostname match (not a domainMatch as per usual).

.setCookieSync(cookieOrString, currentUrl[, options])

Synchronous version of setCookie; only works with synchronous stores (that is, the default MemoryCookieStore).

.getCookies(currentUrl[, options][, callback(err, cookies)])

Retrieve the list of cookies that can be sent in a Cookie header for the current URL. Returns a promise if a callback is not provided.

Returns an array of Cookie objects, sorted by default using cookieCompare.

If an error is encountered it's passed as err to the callback, otherwise an array of Cookie objects is passed. The array is sorted with cookieCompare() unless the {sort:false} option is given.

The options object can be omitted and can have the following properties:

• http - boolean - default true - indicates if this is an HTTP or non-HTTP API. Affects HttpOnly cookies.
• secure - boolean - autodetect from URL - indicates if this is a "Secure" API. If the currentUrl starts with https: or wss: then this is defaulted to true, otherwise false.
• now - Date - default new Date() - what to use for the creation or access time of cookies
• expire - boolean - default true - perform expiry-time checking of cookies and asynchronously remove expired cookies from the store. Using false returns expired cookies and does not remove them from the store (which is potentially useful for replaying Set-Cookie headers).
• allPaths - boolean - default false - if true, do not scope cookies by path. The default uses RFC-compliant path scoping. Note: may not be supported by the underlying store (the default MemoryCookieStore supports it).
• sameSiteContext - string - default unset - Set this to 'none', 'lax', or 'strict' to enforce SameSite cookies upon retrieval. See SameSite Cookies below.
• sort - boolean - whether to sort the list of cookies.

The .lastAccessed property of the returned cookies will have been updated.

.getCookiesSync(currentUrl, [{options}])

Synchronous version of getCookies; only works with synchronous stores (for example, the default MemoryCookieStore).

.getCookieString(...)

Accepts the same options as .getCookies() but returns a string suitable for a Cookie header rather than an Array.

.getCookieStringSync(...)

Synchronous version of getCookieString; only works with synchronous stores (for example, the default MemoryCookieStore).

.getSetCookieStrings(...)

Returns an array of strings suitable for Set-Cookie headers. Accepts the same options as .getCookies(). Simply maps the cookie array via .toString().

.getSetCookieStringsSync(...)

Synchronous version of getSetCookieStrings; only works with synchronous stores (for example, the default MemoryCookieStore).

.serialize([callback(err, serializedObject)])

Returns a promise if a callback is not provided.

Serialize the Jar if the underlying store supports .getAllCookies.

NOTE: Custom Cookie properties are discarded. If you want a property to be serialized, add the property name to the Cookie.serializableProperties Array.

See Serialization Format.

.serializeSync()

Synchronous version of serialize; only works with synchronous stores (for example, the default MemoryCookieStore).

.toJSON()

Alias of .serializeSync() for the convenience of JSON.stringify(cookiejar).

CookieJar.deserialize(serialized[, store][, callback(err, object)])

A new Jar is created and the serialized Cookies are added to the underlying store. Each Cookie is added via store.putCookie in the order in which they appear in the serialization. A promise is returned if a callback is not provided.

The store argument is optional, but should be an instance of Store. By default, a new instance of MemoryCookieStore is created.

As a convenience, if serialized is a string, it is passed through JSON.parse first.

CookieJar.deserializeSync(serialized[, store])

Sync version of .deserialize; only works with synchronous stores (for example, the default MemoryCookieStore).

CookieJar.fromJSON(string)

Alias of .deserializeSync to provide consistency with Cookie.fromJSON().

.clone([store][, callback(err, cloned))

Produces a deep clone of this jar. Modifications to the original do not affect the clone, and vice versa. Returns a promise if a callback is not provided.

The store argument is optional, but should be an instance of Store. By default, a new instance of MemoryCookieStore is created. Transferring between store types is supported so long as the source implements .getAllCookies() and the destination implements .putCookie().

.cloneSync([store])

Synchronous version of .clone, returning a new CookieJar instance.

The store argument is optional, but must be a synchronous Store instance if specified. If not passed, a new instance of MemoryCookieStore is used.

The source and destination must both be synchronous Stores. If one or both stores are asynchronous, use .clone instead. Recall that MemoryCookieStore supports both synchronous and asynchronous API calls.

.removeAllCookies([callback(err)])

Removes all cookies from the jar. Returns a promise if a callback is not provided.

This is a new backwards-compatible feature of tough-cookie version 2.5, so not all Stores will implement it efficiently. For Stores that do not implement removeAllCookies, the fallback is to call removeCookie after getAllCookies. If getAllCookies fails or isn't implemented in the Store, that error is returned. If one or more of the removeCookie calls fail, only the first error is returned.

.removeAllCookiesSync()

Sync version of .removeAllCookies(); only works with synchronous stores (for example, the default MemoryCookieStore).

Store

Base class for CookieJar stores. Available as tough.Store.

Store API

The storage model for each CookieJar instance can be replaced with a custom implementation. The default is MemoryCookieStore which can be found in lib/memstore.js. The API uses continuation-passing-style to allow for asynchronous stores.

Stores should inherit from the base Store class, which is available as a top-level export.

Stores are asynchronous by default, but if store.synchronous is set to true, then the *Sync methods of the containing CookieJar can be used.

All domain parameters are normalized before calling.

The Cookie store must have all of the following methods. Note that asynchronous implementations must support callback parameters.

store.findCookie(domain, path, key, callback(err, cookie))

Retrieve a cookie with the given domain, path, and key (name). The RFC maintains that exactly one of these cookies should exist in a store. If the store is using versioning, this means that the latest or newest such cookie should be returned.

Callback takes an error and the resulting Cookie object. If no cookie is found then null MUST be passed instead (that is, not an error).

store.findCookies(domain, path, allowSpecialUseDomain, callback(err, cookies))

Locates cookies matching the given domain and path. This is most often called in the context of cookiejar.getCookies().

If no cookies are found, the callback MUST be passed an empty array.

The resulting list is checked for applicability to the current request according to the RFC (domain-match, path-match, http-only-flag, secure-flag, expiry, and so on), so it's OK to use an optimistic search algorithm when implementing this method. However, the search algorithm used SHOULD try to find cookies that domainMatch() the domain and pathMatch() the path in order to limit the amount of checking that needs to be done.

As of version 0.9.12, the allPaths option to cookiejar.getCookies() above causes the path here to be null. If the path is null, path-matching MUST NOT be performed (that is, domain-matching only).

store.putCookie(cookie, callback(err))

Adds a new cookie to the store. The implementation SHOULD replace any existing cookie with the same .domain, .path, and .key properties. Depending on the nature of the implementation, it's possible that between the call to fetchCookie and putCookie that a duplicate putCookie can occur.

The cookie object MUST NOT be modified; as the caller has already updated the .creation and .lastAccessed properties.

Pass an error if the cookie cannot be stored.

store.updateCookie(oldCookie, newCookie, callback(err))

Update an existing cookie. The implementation MUST update the .value for a cookie with the same domain, .path, and .key. The implementation SHOULD check that the old value in the store is equivalent to oldCookie - how the conflict is resolved is up to the store.

The .lastAccessed property is always different between the two objects (to the precision possible via JavaScript's clock). Both .creation and .creationIndex are guaranteed to be the same. Stores MAY ignore or defer the .lastAccessed change at the cost of affecting how cookies are selected for automatic deletion (for example, least-recently-used, which is up to the store to implement).

Stores may wish to optimize changing the .value of the cookie in the store versus storing a new cookie. If the implementation doesn't define this method, a stub that calls putCookie is added to the store object.

The newCookie and oldCookie objects MUST NOT be modified.

Pass an error if the newCookie cannot be stored.

store.removeCookie(domain, path, key, callback(err))

Remove a cookie from the store (see notes on findCookie about the uniqueness constraint).

The implementation MUST NOT pass an error if the cookie doesn't exist, and only pass an error due to the failure to remove an existing cookie.

store.removeCookies(domain, path, callback(err))

Removes matching cookies from the store. The path parameter is optional and if missing, means all paths in a domain should be removed.

Pass an error ONLY if removing any existing cookies failed.

store.removeAllCookies(callback(err))

Optional. Removes all cookies from the store.

Pass an error if one or more cookies can't be removed.

store.getAllCookies(callback(err, cookies))

Optional. Produces an Array of all cookies during jar.serialize(). The items in the array can be true Cookie objects or generic Objects with the Serialization Format data structure.

Cookies SHOULD be returned in creation order to preserve sorting via compareCookie(). For reference, MemoryCookieStore sorts by .creationIndex since it uses true Cookie objects internally. If you don't return the cookies in creation order, they'll still be sorted by creation time, but this only has a precision of 1-ms. See cookieCompare for more detail.

Pass an error if retrieval fails.

Note: Not all Stores can implement this due to technical limitations, so it is optional.

MemoryCookieStore

Inherits from Store.

A just-in-memory CookieJar synchronous store implementation, used by default. Despite being a synchronous implementation, it's usable with both the synchronous and asynchronous forms of the CookieJar API. Supports serialization, getAllCookies, and removeAllCookies.

Community Cookie Stores

These are some Store implementations authored and maintained by the community. They aren't official and we don't vouch for them but you may be interested to have a look:

• db-cookie-store: SQL including SQLite-based databases
• file-cookie-store: Netscape cookie file format on disk
• redis-cookie-store: Redis
• tough-cookie-filestore: JSON on disk
• tough-cookie-web-storage-store: DOM localStorage and sessionStorage

Serialization Format

NOTE: If you want to have custom Cookie properties serialized, add the property name to Cookie.serializableProperties.

  {
    // The version of tough-cookie that serialized this jar.
    version: 'tough-cookie@1.x.y',

    // add the store type, to make humans happy:
    storeType: 'MemoryCookieStore',

    // CookieJar configuration:
    rejectPublicSuffixes: true,
    // ... future items go here

    // Gets filled from jar.store.getAllCookies():
    cookies: [
      {
        key: 'string',
        value: 'string',
        // ...
        /* other Cookie.serializableProperties go here */
      }
    ]
  }

RFC 6265bis

Support for RFC 6265bis revision 02 is being developed. Since this is a bit of an omnibus revision to the RFC 6252, support is broken up into the functional areas.

Leave Secure Cookies Alone

Not yet supported.

This change makes it so that if a cookie is sent from the server to the client with a Secure attribute, the channel must also be secure or the cookie is ignored.

SameSite Cookies

Supported.

This change makes it possible for servers, and supporting clients, to mitigate certain types of CSRF attacks by disallowing SameSite cookies from being sent cross-origin.

On the Cookie object itself, you can get or set the .sameSite attribute, which is serialized into the SameSite= cookie attribute. When unset or undefined, no SameSite= attribute is serialized. The valid values of this attribute are 'none', 'lax', or 'strict'. Other values are serialized as-is.

When parsing cookies with a SameSite cookie attribute, values other than 'lax' or 'strict' are parsed as 'none'. For example, SomeCookie=SomeValue; SameSite=garbage parses so that cookie.sameSite === 'none'.

In order to support SameSite cookies, you must provide a sameSiteContext option to both setCookie and getCookies. Valid values for this option are just like for the Cookie object, but have particular meanings:

1. 'strict' mode - If the request is on the same "site for cookies" (see the RFC draft for more information), pass this option to add a layer of defense against CSRF.
2. 'lax' mode - If the request is from another site, but is directly because of navigation by the user, such as, <link type=prefetch> or <a href="...">, pass sameSiteContext: 'lax'.
3. 'none' - Otherwise, pass sameSiteContext: 'none' (this indicates a cross-origin request).
4. unset/undefined - SameSite is not be enforced! This can be a valid use-case for when CSRF isn't in the threat model of the system being built.

It is highly recommended that you read RFC 6265bis for fine details on SameSite cookies. In particular Section 8.8 discusses security considerations and defense in depth.

Cookie Prefixes

Supported.

Cookie prefixes are a way to indicate that a given cookie was set with a set of attributes simply by inspecting the first few characters of the cookie's name.

Cookie prefixes are defined in Section 4.1.3 of 6265bis.

Two prefixes are defined:

1. "__Secure-" Prefix: If a cookie's name begins with a case-sensitive match for the string "__Secure-", then the cookie was set with a "Secure" attribute.
2. "__Host-" Prefix: If a cookie's name begins with a case-sensitive match for the string "__Host-", then the cookie was set with a "Secure" attribute, a "Path" attribute with a value of "/", and no "Domain" attribute.

If prefixSecurity is enabled for CookieJar, then cookies that match the prefixes defined above but do not obey the attribute restrictions are not added.

You can define this functionality by passing in the prefixSecurity option to CookieJar. It can be one of 3 values:

1. silent: Enable cookie prefix checking but silently fail to add the cookie if conditions are not met. Default.
2. strict: Enable cookie prefix checking and error out if conditions are not met.
3. unsafe-disabled: Disable cookie prefix checking.

Note that if ignoreError is passed in as true then the error is silent regardless of the prefixSecurity option (assuming it's enabled).

Copyright and License

BSD-3-Clause:

 Copyright (c) 2015, Salesforce.com, Inc.
 All rights reserved.

 Redistribution and use in source and binary forms, with or without
 modification, are permitted provided that the following conditions are met:

 1. Redistributions of source code must retain the above copyright notice,
 this list of conditions and the following disclaimer.

 2. Redistributions in binary form must reproduce the above copyright notice,
 this list of conditions and the following disclaimer in the documentation
 and/or other materials provided with the distribution.

 3. Neither the name of Salesforce.com nor the names of its contributors may
 be used to endorse or promote products derived from this software without
 specific prior written permission.

 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
 POSSIBILITY OF SUCH DAMAGE.