A Cost Mode Registry for the Application-Layer Traffic Optimization (ALTO) Protocol

The cost mode attribute indicates how costs should be interpreted when communicated in the Application-Layer Traffic Optimization (ALTO) Protocol . The base ALTO specification includes a provision for only two modes: Indicates that numerical operations can be performed (e.g., normalization) on the returned costs (Section 6.1.2.1 of ). Indicates that the cost values in a cost map represent ranking (relative to all other values in a cost map), not actual costs (Section 6.1.2.2 of ). Additional cost modes are required for specific ALTO deployment cases (e.g., ). In order to allow for such use cases, this document relaxes the constraint imposed by the base ALTO specification on allowed cost modes () and creates a new ALTO registry to track new cost modes ().

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 BCP 14 when, and only when, they appear in all capitals, as shown here. This document makes use of the terms defined in .

This document updates Section 6.1.2 of as follows: OLD: The cost mode attribute indicates how costs should be interpreted. Specifically, the cost mode attribute indicates whether returned costs should be interpreted as numerical values or ordinal rankings. It is important to communicate such information to ALTO clients, as certain operations may not be valid on certain costs returned by an ALTO server. For example, it is possible for an ALTO server to return a set of IP addresses with costs indicating a ranking of the IP addresses. Arithmetic operations that would make sense for numerical values, do not make sense for ordinal rankings. ALTO clients may handle such costs differently. Cost modes are indicated in protocol messages as strings. NEW: The cost mode attribute indicates how costs should be interpreted. This document defines two cost modes (numerical values and ordinal rankings), but additional cost modes can be defined in the future. It is important to communicate such information to ALTO clients, as certain operations may not be valid on certain costs returned by an ALTO server. For example, it is possible for an ALTO server to return a set of IP addresses with costs indicating a ranking of the IP addresses. Arithmetic operations that would make sense for numerical values, do not make sense for ordinal rankings. ALTO clients may handle such costs differently. Cost modes are indicated in protocol messages as strings. Future documents that define a new cost mode SHOULD indicate whether that new cost mode applies to all or a subset of cost metrics. If not explicitly indicated, the new cost mode applies to all cost metrics.

This document updates Section 10.5 of as follows: OLD: A cost mode is encoded as a string. The string MUST have a value of either "numerical" or "ordinal". NEW: A cost mode is encoded as a string. The string MUST be no more than 32 characters, and it MUST NOT contain characters other than US-ASCII alphanumeric characters (U+0030-U+0039, U+0041-U+005A, and U+0061-U+007A), the hyphen-minus ('-', U+002D), the colon (':', U+003A), or the low line ('_', U+005F). Cost modes reserved for Private Use are prefixed with "priv:" (). Otherwise, the cost mode MUST have a value that is listed in the registry created in of RFCXXXX.

This document requests IANA to create a new subregistry entitled "ALTO Cost Modes" under the "Application-Layer Traffic Optimization (ALTO) Protocol" registry available at . The registry is initially populated with the following values:

The assignment policy for this registry is "IETF Review" (Section 4.8 of ). Cost modes prefixed with "priv:" are reserved for Private Use (Section 4.1 of ).

This document does not introduce new concerns other than those already discussed in Section 15 of .

Many thanks to Benjamin Kaduk for spotting the issue during the review of . Thanks to Adrian Farrel, Dhruv Dhody, Luis Miguel Contreras Murillo, Sabine Randriamasy, and Qiao Xiang for the review and comments. Special thanks to Kai Gao for Shepherding the document. Thanks to Martin Duke for the AD review. Thanks to Roni Even for the gen-art review and Jaime Jimenez for the artart review.