IS-IS Extension for Big TLV

Type-Length-Value (TLV) encoding of information is widely used in Intermediate System to Intermediate System (IS-IS) routing protocol messages including Link State Protocol Data Units (LSPs). Each TLV defined in allows for maximum of 255 octets of value (or say payload). This is because the length field of the TLV is one octet, which has 255 as its maximum value. When the size of the value of a TLV of type T (such as the Extended IS Reachability TLV of type 22) is bigger than 255 octets, this TLV is called a Big TLV of type T (or Big-TLV for short). There is no general mechanism for encoding and distributing this Big-TLV in classic IS-IS. IS-IS has been optionally extended by which permits larger TLV value, in principle up to 65,535 octets due to a two-octet length field. However, the extensions are not widely deployed, are not backward compatible in the sense that they use a new Protocol Data Unit (PDU) and new LSP types that un-extended implementations will ignore, and in any case do not support values so large they do not fit into a single packet. This document proposes a simple IS-IS extension for encoding and distributing the Big-TLV whose value parts are bigger and can't be accommodated in a single TLV. This extension uses a "Container TLV".

This document describes a method that segments an original Big-TLV into fragmented contents, encapsulates the fragmented contents into one newly defined container TLV, which is transmitted via normal IS-IS PDU to the receiver. The fragment procedure happen only when the length of original Big-TLV exceeds the normal boundary value(255 bytes), and the results segmented contents should be encapsulated within the newly defined container TLV. The original Big-TLV includes sub-TLVs, or sub-sub-TLVs. The segment process occurs only at the boundary of sub-TLV, or the boundary of sub-sub-TLV. That is to say, one normal sub-TLV, or sub-sub-TLV shouldn't be splitted into several segments. The container TLV is defined in , which contains a length of fragmented content(normally less than 255 bytes); a packet identifier to identify the original Big-TLV that the fragmented content belongs to; a packet length of the original data packet; a flag field indicating whether the original Big-TLV includes the sub-TLV and/or the sub-sub TLV and the TLV/sub-TLV and sub-sub TLV types of original Big-TLV. When the receiver receives the segmented container TLV, it will concatenate these container TLVs to reconstruct the orginal Big TLV. The receiver concatenates these container TLVs only when the above mentioned packet identifier, TLV type, sub-TLV type and sub-sub-TLV type are the same. The concatenates procedure will end when the receiver receives all the segmented contents, that is to say, the length of concatenate packet match the original packet length that encapsulated in the header of the container TLV.

A new TLV, called the Container TLV, is defined. shows the format of the new TLV in the classic case. This new TLV is used to carry a piece of the value of a Big-TLV of type T.

Type (TBD1) field: 1 octet. The type of the Container TLV, its value is assigned by IANA. Length field: 1 octet. The length of the Value field of the Container TLV. Value field: contains a Flag field, a Type T0, T1 and T2 field (indicates the top TLV, sub-TLV, sub-sub-TLV type of original Big TLV belongs to respectively), Length of original Type T0/T1/T2, a Identification field and a Piece of value of Big TLV. (Piece field for short). Flag field: 1 octet. Defines flag for the container TLV. T1(bit 7), T2(bit 6) are defined now. when T1/T2 are all unset, it indicates the Big-TLV belong to one top IS-IS TLV; when only T1 bit is set, it indicates the Big-TLV belong to sub-TLV; when both T1 and T2 bit are set, i indicates the Big-TLV belongs to sub-sub-TLV. T2 bit can be set only when T1 bit is set at the same time. Type T0、T1、T2 fields: All are 1 octet, Indicates the Type/sub-Type/sub-sub-Type of the Big-TLV that is being transported in this Container TLV. Length of Type T0/T1/T2: 2 octets. Actual lengths of Big-TLV. Identification field: 2 octets. Together with Type(T0/T1/T2), can identify unique Big-TLV that each container TLV belongs to. The sender of Big TLV should keep the identification filed same when it fragments the Big-TLV into several pieces which encapsulated within the container TLV. The receiver of the Big-TLV should assemble the container TLV with the same identification field and Type(T0/T1/T2) into the original Big-TLV. Piece field: A piece of the value of the Big- TLV of type T0/T1/T2 that is being transported in this Container TLV. When a node has a Big-TLV of type T0/T1/T2 to be originated, it splits the value of the Big-TLV into a number of pieces, from Piece 1 to Piece n. Each piece from Piece 1 to Piece n is less than 247 octets. This is illustrated in .

Each piece carries a subset of entries in the Big-TLV. An entry is an existing sub-TLV or structure. One entry MUST NOT be split across pieces. The Big-TLV is split in entries boundaries, not octets boundaries. The node originates n TLVs for the Big-TLV of type T. These TLVs are the n new TLVs of type TBD1, each of which has a normal payload. The node advertises each of these TLVs to its neighbors according to the normal IS-IS procedure. shows the encoding of the Big TLV with type T in .

For each of the n pieces of the value of the Big-TLV, a Container TLV of type TBD1 carries the piece. Container TLV 1 contains Piece 1 of the value of the Big-TLV; Container TLV 2 contains Piece 2 of the value of the Big-TLV; ...; Container TLV n contains Piece n of the value of the Big-TLV. If a node supports the extension (i.e., Container TLV), the node understands each piece of the value of the Big-TLV received. Each of the n Container TLVs having Type (T) contains a piece of the Big-TLV value in its Piece field.

This section discusses a couple of ways in which a Big-TLV is split into pieces at sending and the pieces are glued at receiving. When a TLV of type T is too big at an originating node, this Big-TLV is split into a sequence of pieces. Each piece carries a subset of entries in the Big-TLV. An entry is an existing Sub-TLV or structure. if there is only one Big-TLV of type T, the node originates container TLVs with type T containing the pieces, and set the identification field the same value in each container TLV. When there are multiple Big-TLVs of type T, the node originates multiple sequences of container TLVs, with each sequence of the container TLVs has unique different identification field. For example, suppose that a node has a Big-TLV of type T = 22 as shown in . This TLV is too Big and split into two pieces piece 1 and piece 2 at boundary between Sub-TLV K and K+1.

For this Big-TLV of type T = 22, the node originates two container TLV with type T = 22 containing the two pieces (i.e., piece 1 and piece 2) directly. The container TLV is illustrated in .

After receiving the container TLV with type T = 22 and the same identification filed(ID X), the node can glue the piece 1 and piece 2 directly accordingly to the F flag that indicate the first first piece and other sequences piece. Alternatively, when a node has multiple Big-TLVs of type T=22, for each Big-TLV of type T split into a sequence of pieces, the node originates a sequence of container TLVs with type T and unique identification value for each sequence. After receiving the container TLV with type T = 22, and different sequence of container TLVs, the node can glue the piece 1 and piece 2 for each sequence through the same identification value. The sliced container TLVs of one Big-TLV SHOULD be encapsulated within one LSP if all the lengths of the container TLV can fit into one LSP. If such condition can't be met, these sliced container TLVs can be put into different LSPs. If the container TLVs of one Big-TLV locate in different LSPs, the receiver SHOULD wait until it receives all these updated LSPs, then begin the concatenating process and SPF calculation. Such process is similar as the general incremental updates of LSPs.

A new sub-TLV, called Big-TLV Capability sub-TLV, is defined in the Router Capability TLV . A node advertising this sub-TLV indicates that the node supports the Big-TLV. The format of the sub-TLV is shown in .

Type (TBD2) field: 1 octet. The type of the Big-TLV Capability sub-TLV, its value is assigned by IANA. Length field: 1 octet. Its value is 1. Flags field: 1 octet. The field of flags. No flag is defined now. A node supporting the Big-TLV MUST advertise this sub-TLV in a Router Capability TLV.

For a network using IS-IS, users can deploy the extension for Big-TLV in a part of the network step by step. The network has some nodes supporting the extension (or say new nodes for short) and the other nodes not supporting the extension (or say old nodes for short) before the extension is deployed in the entire network. The pieces of the Big-TLV, advertised in the Container TLVs, will only be understood by the new nodes and will be ignored by the old nodes. The originator of the Big-TLV MUST consider the above properties when splitting the Big-TLV into multiple pieces. If all the nodes need to have the same new information for using the new information, every node needs to check if all the nodes support the Big-TLV capability which is distributed by the nodes supporting it. If all the nodes support it, every node uses the new information. If it is not required that all the nodes must have the same new information for using the new information, the nodes supporting the Big-TLV capability can use the new information, the nodes not supporting the Big-TLV capability ignore the new information.

The mechanism described in this document does not raise any new security issues for the IS-IS protocols.

TBD

IANA is requested to make a new allocation in the "IS-IS TLV Codepoint Registry" under the registry name "IS-IS TLV Codepoints" as follows:

IANA is requested to make a new allocation under the registry name "IS-IS Sub-TLVs for IS-IS Router CAPABILITY TLV" as follows: