From Erights

Revision as of 04:35, 16 May 2009 by Kevin Reid (Talk | contribs)
Jump to: navigation, search

See [1] for description and protocol documentation. The material should be moved here and rewritten into a specification.

Links to specification details should go right here.

Specification exits

The CapTP specification, per se, does not specify these choices:

  1. The set of Miranda messages all objects exposed over CapTP are expected to respond to. CapTP requires at least __whenMoreResolved.
  2. The format in which CapTP messages are serialized for transmission over a byte stream.
    1. The format in which objects embedded in CapTP messages are serialized in that stream.
      1. The set of graph exits; that is, those objects/makers which both ends know about and can refer to in their serialized objects. The ad-hoc serialization comity proposal would mitigate this.

In an E system, these holes are specified to be filled with, respectively:

  1. Miranda protocol
  2. XXX Write this specification. See CapTP impl in E-on-CL
    1. Data-E, with CapTP-specific details as below.

Data-E configuration

CapTP as fully implemented in E-on-Java uses Java serialization. The work-in-progress implementation of CapTP in E uses Data-E to serialize CapTP message arguments:

(This has all been made up by User:Kevin Reid; please critique.)

The graph exits which are CapTP-specific are:

The uncall recipient for all descriptors. For example, an ImportDesc is represented in Data-E as CapTP_1_descs.import(importPos).
The local presence of the LocatorUnum.

XXX Specify exactly what other graph exits should be available. for now it's the subgraph kit's default scope plus Ref and import__uriGetter.

Personal tools
more tools