OpenLCB: Abbreviated Default Configuration Description Information

This is a working note in which we are developing recommendations for “Abbreviated Configuration Description Information” (ACDI).

There are separate documents that discuss (a) Configuration Protocol and the full mechanisms for Configuration Description Information. In this document, we discuss an abbreviated form of configuration description information that is useful for rapid retrieval of basic information from an OpenLCB node.

Consider a "network browser" that can show more specific information on the nodes, e.g. their type information, user-given names, etc. To do that, the browser could

That last step is very fast. The browser could do the necessary configuration reads in a small fraction of a second, at least on an idle CAN segment.

But the first two steps are not-so-fast. The CDI is a very powerful and general method to describe all about the node, but it's a handful to read & parse when you just want to get some specific information.

This note discuses a code-free approach that would allow a node implementor to say "I keep my basic information in a common place", so that e.g. browser tools can pick that up quickly. Something like:

1) A node denotes that it has this capability using a specific protocol bit (1 CAN exchange to check that)

2) If it does, the basic information can be found in a specific place in the configuration space (specific, well-known address space, location, and format), so can be directly read via configuration reads (datagram access, with small CAN overhead)

For this to work, there needs to be a common content (so that we don't need another layer of CDI-like stuff to define this content too). Hence this note. What should be in that content?

Some fixed info:

*) A version number, so we can extend this later

*) A manufacturer name

*) A manufacturer-issued node type name

*) Node version? (Covers hardware and software versions)

Some variable info, which the user can change:

*) Name

*) Location

(or a more general comment field? Or both?) These variable items don't have to be provided, but if they are, they can be just the usual identification items that are accessible by the full configuration protocol, configured by the user with the full configuration protocol, but then made available for a quick read this way.

It would be nice to keep this somewhat compact, with the most useful (in the sense of "people will want to see this quickly") info at the front. Reading a full datagram (64 bytes payload originally, but now less based on extended CAN) would let us sample about 50 nodes in a second (idle bus), which is fine. Another factor of 4 larger might start to seem slow to the user, though. The whole idea is to be able to get a view of even large networks faster than a user can browse through the info on an screen. The user can then select a small number of specific nodes that are interesting, pull the full CDI and work with the entire configuration information of those nodes.

Proposed content:

Read from space N (suggest 0xFC, right below CDI etc)

Byte 0: ADCDI version, default 0

Next: Manufacturer name as string

Next: Manufacturer-issued node type name as string

Next: Manufacturer-defined node version, as string

Read from space M (suggest 0xFB, right below CDI etc)

Byte 0: ADCDI version, default 0

Next: User provided node name as string

Next: User provided free-form comment as string

(Comment provides a longer field, not computer parsed; node name is a short thing for e.g. headings, etc)

(String means zero terminated)

Two spaces provided for simplicity in implementation, as we imagine that some will be from a permanent memory and some will be user configurable.

Strings just concatenate with null termination. A missing value is then just a null byte.

Could get in e.g. 30 character manufacturer, 16 character node type, 8 character version into single datagram. (This is just fixed data in memory)

Protocol ID protocol specifies whether this is present, so no more confirmation information is needed.

Later versions with higher ADCDI version numbers could contain more fields, result in more than one datagram in reply, etc. But for expansion purposes, they should be proper supersets of this.

Simple Node Identification Information

(Should this be moved to another doc?)

There is a need for something for tiny nodes, that uses a single message instead of datagrams and configuration. This allows getting getting the information with a bare minimum of support at the requested node end, so long as the requesting node can be sure to handle it.

Upon receipt of an addressed messages with a SNII Request MTI, the node returns a single message with a SNII Reply MTI carrying the information of the two types listed above, concatenated.

On CAN, this is done via the usual mechanism of concatenated frames, with a short or empty one indicating end.

Paragraphs for TN

this particular area is intended to be vague and hand wavy, based on the discussions in the NMRA DCC group. Manufacturers _don't_ want to be told what to call their nodes. They want a place for a "Manufacturer-issued node type" where they can put whatever "node type" name they want to. No more normalization of that content than that is likely to succeed. Certainly not model number; most aren't numbers but some are. Some DCC decoders don't even have model names.

Site hosted by

This is SVN $Revision: 1968 $