Electrical computers and digital processing systems: interprogra – Data transfer between operating systems
Reexamination Certificate
1999-08-03
2004-10-05
Bullock, Jr., Lewis A. (Department: 2126)
Electrical computers and digital processing systems: interprogra
Data transfer between operating systems
C709S245000, C709S238000, C709S230000, C709S227000, C719S313000
Reexamination Certificate
active
06802068
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Technical Field
This invention pertains to computer communication and information networking. More particularly, it concerns generic application level connectivity without depending on an end-to-end network address space.
2. Description of Background Art
Prior networking art embodies a long standing perception that even logical, or application level, connections must be determined by uniquely identifying the physical end points, ie. by a globally unique addresses. The belief is central to the Internet Protocol (IP) suite and is enforced by almost all network application programming interfaces (APIs), including the Berkeley sockets. A symptom of this approach is that the application end points are directly exposed, in the form of (IP_address, port_number) tuples, allowing room for inadvertent or malicious connections, unless protected by a firewall; applications must defend themselves by validating the protocol (eg. HTTP), using magic cookies (eg. the X protocol), or by encryption (eg. SSL). It would also improve security somewhat if the identity of the final destination were concealed or eliminated from the data packets, because anonymous data is often less useful.
More importantly, an IP address is merely a symbolic substitute for the network end point, which means that, notwithstanding its role as an inter-networking architecture, IP does not really solve the problem of confinement by network address boundaries, but works around it by emulating a global end-to-end virtual network: IP applications can run across heterogeneous component networks, but only if the end-point physical hosts (or network interfaces) bear unique IP addresses. This was a strength in the early days of the Internet, because fixing the basic transport format first was crucial to the collaborative development and deployment of the infrastructure protocols of the IP suite. Nevertheless, it has left IP inherently restricted to a finite address space, which means one must currently resort to embedding techniques, such as tunnelling and Network Address Translation (NAT), to extend the existing infrastructure. The restriction more generally means that the associated data structures and protocols must be hardcoded into application software or middleware, and it limits the flexibility and power available to these programs in ways that will be become clear from the description of the present invention.
Accordingly, it is important to avoid application-level dependance on addressing, but because of its crucial role in the development of the Internet, IP-like addressing is commonly assumed to be both necessary, in that any sound scheme for internetworking is expected to critically depend on end-to-end fixed length addresses, andsufficienteven for future internetworking frameworks and applications. These notions are reflected in the view, formalised by the ISO open systems interconnection(OSI) seven-layer model, that the transport mechanism must be solved independently of the other layers,without any help from the latter. This is an undue restriction and a very wrong assumption, as will be particularly shown by the description of the present invention, which exploits simple techniques from the client-server model and operating system (OS) and compilation domains to solve the transport problem in fundamentally different way. The assumption is manifest in the existing specifications of switched networks, including X.25 and Asynchronous Transfer Mode (ATM), which require the signalling to depend on preassigned globally unique multi-byte addresses for the individual switches and host interface adapters. The addressing is currently needed to enable application processes to identify the final destinations without intimate knowledge of the network configuration, only because the current frameworks, such as ATM's network-network interface (NNI) and user-network interface (UNI), were once again conceived without considering higher layer techniques. The result is double addressing and signalling when ATM is used as transport under the IP suite, once for establishing the virtual circuits and once more for emulating IP subnets and virtual LANs (VLANs). Surprisingly, the two-level approach of the present invention, involving connection-oriented networking even over IP, manages to eliminate this duplication.
Another duplication of function concerns the name service. In the earlier Unix-to-Unix Copy (UUCP) system, client applications were required to identify successive hosts all the way to the destination, which put the burden of route discovery and specification squarely on the clients and made the system quite unscalable. The Domain Name Service (DNS) name strings still trace out a logical path to the destination through the DNS hierarchy, but IP goes to the other extreme of not using this logical path structure at all in the routing of data. Instead, IP server applications simply listen at port numbers on their own respective hosts, and their clients are expected to locate them by their host names. This makes the hierarchical organisation of the DNS critical to its operability, as each client's nameserver would otherwise need to be able to locate every server in the IP universe with no geographical hints from the client applications whatsoever.
Also, IP's prescribed use of the addresses for routing is turning out to be inefficient in some ways, and the functionality is now being replaced by Multi-Network Label Switching (MNLS), in which routing labels are affixed to the packets within the network, introducing further duplication into the scheme. Every duplication means avoidable computational or bandwidth overheads, in addition to increased development and maintenance costs. Furthermore, the packet address fields are being extended to implement IPv6, along with the corresponding infrastructure, processing and communication overheads, in every packet, application program, host and router, principally in order to accomodate the growing IP membership. These costs could have been considerably avoided had the Internet not been address space-dependent.
Additionally, per the traditional prescription, the final destination addresses must be interpreted at every router, bridge and gateway along the way. There is no protocol-independent notion of logical connectivity in the IP suite, nor in other address-oriented internetworking suites of the past, so that every logical transport path must be freshly established on a per-packet basis. The difficulty and limitation this imposes is that between any given pair of application end-points, the connectivity must be independently established for each transport stream, and depending on the protocol, may not be possible at all. This is becoming especially clear with the emergence of streaming multimedia applications, where the clients conceptually make logical connections over TCP using HTTP, but the preferred media streams involve RTP over UDP and are stopped by most corporate firewalls. The problem is currently addressed by application-level proxies, but this is a piece-meal approach, as newer protocols are being formulated all the time, and is an impediment to the development of newer network applications. For example, the SOCKS V5 protocol finally supports UDP relay, but applications still need to be specially compiled and linked, or the OS specially SOCKSified, for it to work, and it provides only one-way traversal across a set of firewalls. If the networking were instead inherently connection-oriented, firewall traversals would never have been a special issue, since the authentication could then be applied to the logical connection once, as will be demonstrated by the present invention, for any number of firewalls and transport streams.
In any case, IP addresses are losing their one-time significance as long term identifiers of client hosts, as more and more clients use dial-up connections via Internet Service Providers (ISPs) and even office equipment is migrating to Dynamic IP. A similar trend may be noticed in the server space, as servers of
Bullock, Jr. Lewis A.
F.Chau & Associates LLC
Percello Louis J.
LandOfFree
Addressless internetworking does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Addressless internetworking, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Addressless internetworking will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3290962