Patent Application
20050086295
The present invention relates to electronic messaging methods and more particularly to asynchronous hypertext messaging between clients and servers with hypertext transfer protocol (“HTTP”) capability.
Electronic messaging methods are well known in the art and are an integral part of the suite of computer applications available to users of the internet. A common example of electronic messaging is instant messaging. However, for real-time information and data transmissions, instant messaging has significant drawbacks.
A popular application for instant messaging over the internet is the internet relay chat (“IRC”) application. which is executed at a higher level protocol than the general internet transmission protocol, transmission control protocol/internet protocol (“TCP/IP”). IRC typically requires the client to execute IRC client software and connect to a specified IRC server where client to client real-time communication can occur. IRC client software typically utilizes a data destination port which is reserved only for IRC communications. However, gateway computers implemented for network security reasons, also known as “firewalls,” typically do not enable the IRC port because of network security concerns. Firewalls are designed to limit the data traffic that may enter or leave a network to known protocols. That is, protocols that have been accepted as benign within a defined context are allowed to pass through the firewall to the network, while other protocols are prohibited from entering the client network. With the tremendous growth of the World Wide Web, and its attendant informational potential, the “web browsing” protocol, also known as hypertext transfer protocol (“HTTP”), is generally allowed to pass through firewalls without any modifications. IRC ports, on the other hand, have not generally been enabled on firewalls because of network security concerns.
Therefore, there is a need for an asynchronous real-time data or information delivery system which can use the standard HTTP protocol for the modern networked computing environment without compromising network security.
The present invention utilization of an asynchronous hypertext messaging system provides several major benefits. First, the present invention enables the server to force real-time updateable information, such as pricing information, supplied to the server from a resource, to computer workstations over the internet without polling during a given communication session. This is sent via a pseudo web browser pathway created between these computer workstations in response to an HTTP request for a large document created by active hypertext messaging conversion of messages generated from the computer workstations.
Second, all data transfers (messages) are encoded to appear as HTTP messages. The information is passed transparently through firewalls without the need for network security to be lowered.
Third, at the server end, the server is able to distinguish between the asynchronous hypertext messaging from standard HTTP requests. The server handles both standard web browser requests and asynchronous hypertext messages simultaneously. When a standard web browser request is submitted to the server, the request is forwarded to an available web server.
Fourth, asynchronous hypertext message encoding is efficient. Many logical data streams may be multiplexed onto a single asynchronous hypertext message stream. Optimization of server resources can occur by unloading the server of requirements for multiple simultaneous socket connections.
Finally, asynchronous hypertext messaging enables data encryption as specified by the server encryption components. Therefore, asynchronous hypertext messages can provide a permanent virtual connection via a HTTP proxy server or a secure HTTP proxy server without compromising the security model.
Broadly, in one aspect, this invention is a method of transmitting asynchronous hypertext messages from a server to a client interconnected by an internet, the method comprising the steps of transmitting a HTTP request identifying the client from the client to the server; initiating a virtual connection from the server to the client wherein the server transmits a HTTP header signifying a large HTML document and transmitting asynchronous hypertext messages encoded in the HTML document from the server to the client via the virtual connection wherein the virtual connection is maintained by periodically refreshing the HTML document at the client.
In another aspect, this invention is a trading instrument communication system for communicating trading information over an internet between computer workstations, the communication system comprising means for providing active hypertext messaging trading information between the computer workstations over the internet, the computer workstations being capable of generating messages, the hypertext messaging means comprising means for converting the messages to be transmitted over the internet; and means for transacting trades for designated trading instruments based on the communicated trading information.
In another aspect, this invention is a communication system for communicating real-time updateable information over the internet between computer workstations without polling, the communication system comprising means for providing active hypertext messaging information between the computer workstations over the internet, the computer workstations being capable of generating a real time message, the active hypertext messaging means comprising means for converting the generated message into a pseudo HTTP browser type message for transmission over the internet; server means disposed between the computer workstations and connectable thereto over the internet; and means for supplying the real time updateable information to the server means during a communication session, the server means comprising means for creating a pseudo large document message pathway between at least a pair of the computer workstations in response to the pseudo HTTP browser type message for enabling the real time updateable information supplied to the server means to be communicated to the pair of computer workstations over the pathway without polling during the communication session.
Other features and advantages of this invention will be apparent to those skilled in the art from this disclosure.
A more complete understanding of the present invention may be obtained by considering the following description in conjunction with the drawings in which:
FIGS. 3 is a schematic diagram illustrating the multiplexing of multiple logic streams between objects in accordance with the principles of the present invention;
The asynchronous hypertext messaging system and method of the present invention uses existing hypertext transfer protocols to provide a real-time interactive messaging system which can tunnel through firewalls.
Hypertext transfer protocol (“HTTP”) is a well known application-level protocol for distributed, collaborative, hypermedia information systems. At the heart of HTTP is a generic, stateless request and response protocol controlled typically by web browser software at the client and by web servers at the server. For example, when a web browser implemented on a client requests information from a web server implemented on a remote server, the web server subsequently responds to the request with the requested information.
Referring now to the drawings in detail, and initially to
For purposes of this application, the client 11 is representative of a single terminal, a computer workstation, a personal computer or, more generally, a network of computers. The client may also be representative of a local area network. Similarly, the server is representative of a single web server or a plurality of clustered web servers capable of delivering documents via the conventional HTTP protocol. The client firewall 21 is typically a separate computer which acts as a gateway computer and acts to virtually isolate the users of the clients from the internet or other external networks. For purposes of this application, the firewall is representative of proxy servers or any other conventional network security measures implemented at the perimeter of a network. Within the security perimeter of the firewall, users may be able to communicate freely. However, all messages sent to or from users outside the client must pass through the firewall computer, or set of computers, which will check, route, and frequently label all information that passes through it. A firewall can be a conventional computer running specific firewall software, or a dedicated computer device specifically constructed or configured as a firewall. The firewall can be dedicated solely to performing the firewall functions, or it can also perform additional functions such as packet routing, in addition to its firewall functions. Such firewalls will block data packets according to specific sets of predetermined rules programmed into the firewall.
The server firewall 41 is implemented in similar fashion to the client firewall at the web server 51. Thus, the server firewall 41 is typically a separate computer which acts as a gateway computer and acts to virtually isolate the servers from the internet or other external networks. As with the client firewall 21, within the security perimeter of the firewall, servers may be able to communicate freely. However, all messages sent to or from users outside the server must pass through the firewall computer, or set of computers, which will check, route, and frequently label all information that passes through it. As noted above, a firewall can be a conventional computer running specific firewall software, or a dedicated computer device specifically constructed or configured as a firewall. The firewall can be dedicated solely to performing the firewall functions, or it can also perform additional functions such as packet routing, in addition to its firewall functions. Such firewalls will block data packets according to specific sets of predetermined rules programmed into the firewall. The server firewall 41 typically allows HTTP requests from other components of the internet to pass through transparently to the web server while blocking other types of data packets.
In the conventional system of
However, a typical client firewall 21 is configured so that restrictions are placed on many of the ports used for proper data communication. A HTTP request, however, is considered benign and is generally passed through to the internet. As long as HTTP requests are initiated from the client, resulting HTTP responses are conventionally given a defined dedicated path back to the client. Accordingly, the real-time interactive communication methodology of the present invention encompasses the HTTP protocol in order for the firewalls to be effective in their default configuration.
There is a prior art solution for delivering real-time data to a client within the HTTP protocol although such solution is not satisfactory due to the need for continuous polling. Accordingly, in such a prior art solution, software resident on the web page at the web server may force the client software (such as a browser) to continually poll, or refresh, the HTTP request so that the latest information is delivered to the client. Alternatively, software resident on the client may continually poll the web server. However, these approaches have severe limitations. For instance, data or information to be sent to the client must be classified with certain polling intervals. For example, if the data consists of a securities price which must be transmitted to the client within a five second time frame, it may be necessary for the client to request an update from the server once every second to keep the latency to a minimum. This would generate a huge amount of traffic overhead and become impractical for large scale implementations. The polling interval can range from a few seconds (for example, stock or commodity prices) to hours (for example, the price of durable goods at a wholesale oriented website). On the one extreme, a few second polling interval can generate enormous amounts of data traffic for a network of modest size and can generate an enormous amount of overhead which make it impractical for any substantially large scale implementation of real-time data. At the other extreme, the polling interval can become so long that the data or information can no longer be termed real-time.
A more efficient solution which overcomes the unsatisfactory problems created by polling is illustrated in the present invention. Referring now to
The communication server 61 may be generally implemented as a computer with dedicated software devoted to the delivery of real-time data. The resource servers 81, 82 may be generally implemented as a computer with dedicated software providing the source of real-time data to be routed to clients. The resource servers 81, 82 may be dedicated to delivering various types of real-time data to the client. For example, the resource servers may deliver any real-time updateable information, such as real-time pricing information, such as stock prices, or other real-time data such as foreign exchange quotes, bidder and seller information for capital market transactions, etc. In a preferred embodiment, the resource servers 81, 82 deliver real-time data asynchronously. That is, the resource servers deliver data only if the data has changed, or needs to be refreshed. It should be noted that the various firewalls 21, 41 and 71, are similar in operation to those described above. The web server 51 may be a separate computer dedicated to delivering web pages and is, significantly, connected to the communication server 61.
It can be readily seen from
In practical systems, there are often limitations on facilities that the network can make available to an application. The network facilities include entities that are often required to support communications such as file descriptors and TCP ephemeral ports. By multiplexing several logical communication streams onto a single physical transport stream it is possible to reduce the impact on system resources. In other words, great system benefits are achieved when a single client instance can support multiple client side objects that can communicate over a single physical stream with multiple server side objects that are supported by a single AHM server instance. A practical implementation of this within the asynchronous hypertext messaging system of the present invention is enabled by automatically allocating multiplexing identifiers to the communications logic. Two main identifiers are used. Referring to
It is additionally noted that any HTTP requests transmitted from the client 11 to the communication server 61 that are not associated with asynchronous hypertext messaging are directed to an appropriate web server 51 and resulting web server responses are relayed to the requesting client 11. In an alternative embodiment, relayed requests and responses by the communication server 61 are translated to enable appropriate mapping of the web site universal resource location (“URL”). This is especially useful where external entities expect to receive URLs that match specific constraints.
Referring now to
At step S2 the client requests a java applet download from the web server 51. The communication server 61 translates the HTTP request as a true HTTP request and relays the HTTP request to the web server 51. The web server 51 provides a response to the client web browser which includes the java applet download. In an alternative embodiment, the applet is pre-installed onto the client browser and appropriate commands are conventionally issued to initiate the java applet.
At step S3, the client indicates the desire to log in to a resource server 81, 82 for secure real-time communication. The java applet subsequently issues a “GET” command to the communication server 61 so that the communication server is notified of the intended resource server 81, 82. An example of the HTTP command issued by the java applet is:
At step S4, the first multiplexed virtual channel connection is initiated. A “GET” command is again issued to the communication server 61 so that the communication server may transmit subsequent asynchronous messages to the client 11. An example of the HTTP command issued by the java applet is:
At step S5, the communication server 61 receives a multiplexed channel creation request and returns an HTTP response header. The response header indicates that the communication server 61 is set to transmit that virtual document (mux.html in this case) to the client 11 and that the document is very large. Appendix A contains exemplary code for the communication server 61 to transmit to signal to the client 11 that the document is to be a very large document. This creates a pseudo large document message pathway between the communication server 61 and the computer workstation 11. The response header conventionally readies the web browser at the client 11 for what is expected to be a large document so that the web browser connection is maintained throughout the transmission or communication session. The first few bytes of the response body is sent by the server. An example of the format of the body message to be sent by the communication server is:
At step S6, the client java applet creates a unique ID identifying a specific resource server 81, 82 with a <cID>, a channel identification number. An example of an HTTP message sent from the client java applet is:
At step S7, the resource servers 81, 82 asynchronously send real-time data to the client 11. Real-time data is transmitted to the client along the multiplexed connection in the form of an HTTP message such as:
At step S8, the client java applet receives asynchronous real-time data from the multiplexed connection and passes the data to the logical client indicated by the <cID>. At this point, the java applet may format the real-time data and provide a visual display of the real-time data to the user of the computer workstation 11. Alternatively, the java applet may be conventionally modified to display the real-time data in a scrolling stock ticker display window at the computer workstation 11.
At step S9, the client java applet can asynchronously send real-time data to the server-side resource 81, 82. The client java applet can send a command in the form of:
Typical web browsers feature time-out provisions to close HTTP connections when the web servers do not respond after a prespecified time interval. To prevent the connection from being closed by the web browser software or the firewall 21, 41, 71 due to possible gaps between real-time data transmission from the resource server 81, 82, in step S10, a refresh command is issued by the client java applet. The client java applet issues a command in the form of:
Referring now to Table 1, the above described procedure is detailed with specific implementation details. Table 1 is organized so that column 1 specifies client initiated processes, column 2 specifies server initiated processes and column 3 illustrates specific comments and notes based on each process. It is noted that, for purposes of this application, active hypertext messaging and asynchronous hypertext messaging are interchangeable.
Referring now to
In a preferred embodiment, the communication server 61 implements a “heartbeat” mechanism. It is a common requirement that a system that requires real time communication will also need an early indication if excessive delays are being incurred between sending and receipt of messages. However, a reliable transport protocol such as TCP may continue to attempt to re-send segments of information under failure conditions for several hours without alerting the application that sent the information. Thus, to enable appropriate action to be taken under failure conditions a best implementation of asynchronous hypertext messaging includes a “heartbeat” mechanism. A regular heartbeat message is sent at a configurable interval from the communication server 61 to the client 11. Alternatively, the heartbeat need not be sent if other valid data is sent. If the client 11 does not receive a heartbeat (or optionally some other valid data) after a configurable number of heartbeat intervals then the client consumer is warned. Alternatively, a second configurable threshold of a greater number of heartbeat intervals may be set. If no heartbeats are received during a period equivalent to this second threshold then the physical communication stream is dropped. This second threshold is useful in circumstances where the system designer is sensitive to the possibility of a large number of “stale” connections being used to create a “Denial Of Service” scenario. An alternative refinement of the heartbeat methodology is obtained where the entity that sends the heartbeat includes the set heartbeat interval within the heartbeat message. This allows the entities to automatically synchronize their heartbeat settings, avoiding problems caused by misconfiguration.
Additional reliability can be achieved under certain circumstances by creating multiple physical streams and sending duplicate data. In this alternative embodiment, the sender (either the client 11 or the communication server 61 ) attaches a unique identifier to each message that is sent. The relationship between identifiers on successive messages must be known to both sender and receiver (for example incrementing the identifier by one on each message, within a defined modulus). The receiver then is able to quietly drop messages that have the same identifier as messages that have already been received. It is also necessary for the receiver to be able to reorder received messages into the correct order before forwarding them to the appropriate resource server or consumer.
It is additionally noted that the client 11 and communication server 61 are able to determine several metrics which give an indication of the performance of the physical communication stream. These metrics can be based on a number of parameters including the time required to perform a refresh request, the delay of heartbeats, the variation of the delays from time to time, etc. These metrics may be passed to the consumer or resource server, sent to a system monitoring entity, and used as the basis for automatic route selection. In addition the metrics may be used to trigger changes in the operation of the system such as instigating “multiple send” or “automatic route selection.”
With a system monitoring entity, in certain operational environments there may be several accessible routers available for TCP/IP based communication. In a preferred embodiment, a communication server 61 allows the server or the client to select, in real time, the most appropriate next hop to send messages. The criteria for the routing decision may be based on type of service (such as large non-delay sensitive messages vs. smaller delay sensitive messages) or based upon measurements obtained through real time connection monitoring. This functionality allows intelligent routing to take place based upon HTTP protocol performance as opposed to the underlying TCP/IP performance. A typical application may be where connections are available to two Internet Service Providers (“ISPs”) and that real time HTTP performance to a given customer is significantly better through one ISP than the other. Or if a private wire is dedicated to the connection but a problem occurs that renders it unavailable the information may be temporarily switched to an internet based route.
It must be noted that automatic route selection can still be used where the client is implemented as an unsigned Java applet (thus having the restriction that it can only communicate with the IP address from which it was downloaded). This can be achieved by specifying to the client that it should indicate its route selection by means of destination port rather than IP address. The conventional gateway router can then make a next hop routing decision based upon the specified layer 4 port, rather than the layer 3 IP address. Most TCP/IP capable routers support this functionality.
In regards to security issues for the real-time data, there are a number of methodologies which may be employed to encrypt the data. One methodology is using HTTP over Secure Socket Layer (“HTTPS”). A link will be encrypted using a symmetric key that has been created securely using the trusted asymmetric key in the SSL X509 certificate based methodology. A more secure link can be created by using HTTPS plus additional key Methodology. After the HTTPS link has been created, the communication server creates a new symmetric key on the fly. This key is securely passed to the client using the HTTPS link. Client and Server communications within the session are, from then on, encrypted with the new key. These encrypted messages are then further encrypted as they are passed across the HTTPS channel. An alternative encryption methodology is HTTPS followed by additional key on HTTP. The methodology utilizes the same procedure as for HTTPS plus additional key except that the link operates wholly or partly across an HTTP link rather than an HTTPS one. Thus the AHM generated symmetric key is passed back to the client across an HTTPS connection, but the main communications take place across an HTTP connection encrypted only with the AHM generated key. An alternative encryption methodology is HTTPS followed by additional key on raw TCP/IP socket. This methodology is identical to the aforementioned HTTPS followed by additional key on HTTP, except that after the client has received the server generated key, further communications take place over a standard TCP/IP socket.
With respect to authentication, it is noted that additional authentication, if desired, is possible between the server and the client. Either in conjunction with a simple user id/password scheme, or instead of, the client can indicate that additional authentication scripts must be run. Such a script would normally inspect a previously authenticated client identifying token such as a client certificate or a previously placed secure cookie. As a result of running the script, the communication server would return information such as an mID to identify the physical connection, a user id/password pair, and optionally an additional key to further encrypt the channel.
There are several ways that authentication can be achieved in a practical implementation. In one embodiment, the client is allowed to communicate directly with a secure web server (normally this would be an HTTPS connection). The secure web server obtains the client token and runs the script, returning the results directly to the client. It is necessary for the script to communicate with the server side component to obtain an identifier for the physical stream (the mID).
In a second embodiment, the client communicates only with the server side component incorporated within the communication server. All HTTPS requests/responses that the communication server receives are decrypted/encrypted by the communication server. Therefore, all true HTTPS requests received by the communication server are relayed to/from the associated web server in unencrypted, HTTP form. Thus although communication between the communication server and the client would typically be HTTPS listened for by the communication server on the well known HTTPS port 443, the request for applet download and other true browser requests would typically be relayed on to a web server listening on the well known HTTP port 80. Therefore, the web server will not receive the client's token such as certificate details directly. However the communication server will receive the token details and is able to make those details available to an external process. This could be made available as environment variables, or more efficiently, the communication server posts client token information to the attendant web server as part of the process of relaying on the script that has been indicated by the client. The token information can be appended by the communication server to the HTTP script call as either part of an HTTP GET command or as part of an HTTP POST command. In either case, the script which is run by the web server can easily access this information to perform the required processing. In this way, standard web server scripting/plugin methodologies can be used.
In a third embodiment, the client communicates only with the server side component incorporated within the communication server. The communication server shares a server certificate with the web server. Only HTTPS client requests/responses that are tunneled communications to/from server side resources are decrypted/encrypted by the communication server. Thus all true HTTPS requests received by the communication server are relayed to/from the associated web server in encrypted, HTTPS form. Therefore, the web server will receive the client's token such as certificate details directly. The web server can run the required authentication script and the communication server relays the results back to the client.
It should be noted that a server side asynchronous hypertext messaging implementation encapsulated within a communication server would provide the means to authenticate a user on the basis of the source IP address of the received connection. Resource servers can then expose access control lists based upon IP addresses and IP address ranges that indicate connections that they would allow or deny, possibly pending additional authentication criteria.
In the system of the present invention, the communication server may act as a “proxy consumer” on behalf of remote applets or applications that wish to access server side resources. Generally, security requirements dictate that no remote entity may communicate directly with sensitive server side resources. The proxy consumer functionality provides a secure way of enabling remote objects to consume server side resources. In addition, proxy consumer capability can be achieved without the client consumer requiring any knowledge of the actual location of the server side resources—the client only needs to know the location of the communication server. Further, it is possible for requests from multiple consumers to be load balanced across multiple instances of any server side resource.
Referring now to
Any prospective consumers 11, 91, 92 of resources can receive the benefits of the resource by querying the resource registry 111 and receiving directions on how to access the most appropriate instance of the resource application through the resource API. New resources and consumers can be added via a software library that enables the registering and consuming of resource services.
The process of registering a resource server 81, 82 is achieved as follows. The initialization file of each application contains the location and communication options associated with one or more resource registries 111. On startup an application that wishes to expose a resource registers it by name with the resource registries 111. The registration includes information that forms a “handle” to that resource such as communication details, location, maximum number of connections that can be supported simultaneously, etc. There is no automatic method by which resource registries become known to resources. Resources are entirely reliant on correct configuration within the initialization file to be able to locate a resource registry. Resource to registry connections have a configurable. heartbeat period to monitor the quality of the link. If disconnection of the resource from a resource registry occurs then the link is auto retried. on a configurable interval until reconnection is achieved. Having registered with a resource registry, the resource waits for notifications that a consumer wishes to connect to it. On receiving new consumer connections the resource is required to provide the service which it advertised.
It should be noted that a resource can be running on any machine that has TCP/IP communication with a resource registry. Each resource maintains an open communication stream with each of the resource registries that it has registered with. The resource registry immediately becomes aware if a resource becomes unavailable and will remove that resource's registration so that no further consumer requests are forwarded to it. Under these circumstances the communication server might receive a request for a particular resource by name from a remote consumer.
For example, in a financial trading system, a consumer associated with a remote trader applet that incorporates client side asynchronous hypertext messaging may request a resource known as a deal server. The means by which this request is satisfied might be implemented as follows. The applet connects to the communication server and requests a deal server resource. The communication server asks one of the available resource registries for a handle to the most appropriate available deal server resource. If that resource registry does not know of an available resource of type deal server then the communication server will move on to another resource registry and repeat the request. Given that a handle of an available deal server is returned to the communication server by a resource registry, then the communication server attempts to make a direct connection to that resource. However, if the communication server cannot make a direct connection to that resource, it returns to asking connected resource registries for other instances of the deal server resource. Having established a connection with an instance of the deal server resource, the communication server relays communications between the remote consumer applet and the resource.
It should be noted that further to the above description, a practical implementation of utilizing the communication server as a proxy consumer would generally include local caching of resource registry details to improve real time performance.
It should be noted that in high availability environments it would normally be necessary for multiple instances of communication servers to be available to support client connections. To enable this it is required that a third party IP load balancing solution is logically positioned in front of the communication server instances. Such solutions are generally required to support any high availability web site installation and are designed to make instances of the site on different machines to be perceived to the outside world as all existing on a single machine with a single IP address. This is achieved by IP address translation. In addition the IP load balancer needs to keep track of sessions initiated by specific clients and ensure that once a session is started that packets received in that session continue to be forwarded to the same SCS instance. This is also standard practice for secure web site installations and is generally based upon a timeout scheme associated with packets from each source IP address. As noted above, communication server communications can include an scsID that is unique to a given communication server. This can be used to ensure that the session awareness of an IP load balancing solution is operating correctly.
Although the present invention is described in various illustrative embodiments, it is not intended to limit the invention to the precise embodiments disclosed herein. Accordingly, this description is to be construed as illustrative only. Those who are skilled in this technology can make various alterations and modifications without departing from the scope and spirit of this invention. Therefore, the scope of the present invention shall be defined and protected by the following claims and their equivalents. The exclusive use of all modifications within the scope of the claims is reserved.
| Number | Date | Country | |
|---|---|---|---|
| Parent | 09684309 | Oct 2000 | US |
| Child | 10872130 | Jun 2004 | US |
