Skip to main content
Jump to: navigation, search

Difference between revisions of "Selector Architecture Harmonization"

(Should we use WebDAV in CardSync ?)
(Open Issues)
Line 28: Line 28:
 
Reasons Against:
 
Reasons Against:
 
* Question: Might using protobuf reduce the likelyhood of the CardSpace team collaborating with us and potentially adopting the [[CardSync Protocol]]? <-- an outcome that would increase interoperability in the industry
 
* Question: Might using protobuf reduce the likelyhood of the CardSpace team collaborating with us and potentially adopting the [[CardSync Protocol]]? <-- an outcome that would increase interoperability in the industry
 +
 +
==== LICS design points ====
 +
* Need to have verification/authentication between selector and browser plug-in to LICS. This needs to consider factor in design of LICS.
  
 
== Top Level Architecture==
 
== Top Level Architecture==

Revision as of 23:23, 4 February 2009

{{#eclipseproject:technology.higgins|eclipse_custom_style.css}}

Higgins logo 76Wx100H.jpg

Since Selectors use most of the Higgins Components, work on harmonizing the Higgins selectors into a single architecture would be a huge step towards overall Higgins architecture harmonization/convergence. A good first step in converging the selectors is start by harmonizing the GTK and Cocoa Selector and the AIR Client and Server. Aside from UI differences, the former performs all processing locally, whereas the latter presents a local UI but relies on a hosted server. The latter has the advantage of supporting roaming of cards as well as multiple simultaneous clients. The former has performance advantages. We'd like to get the best of both worlds by having a converged architecture which synchronizes cards between the client and the server. The common code would be in a local "selector service" component that alternative UI layers can use.

This wiki page is being used to coordinate this project to harmonize the architectures of the GTK/Cocoa Selector and the architecture of the AIR Selector.

Open Issues

This is the place to capture open design issues in this project.

CardSync Protocol Open Issues

Should we use WebDAV?

Reasons For:

  • reduces our development effort as this protocol is designed for resource synchronization. And that's what we're doing
  • mature, widely supported. Being used in new projects such as Mozilla Weave.

Reasons Against:

  • Not RESTful. WebDAV extends the HTTP verb set. As such is deprecated by the W3C.
  • Can interfere with web cache architectures

Should we use Google protocol buffers?

Reasons For:

  • Mature
  • High performance
  • Good libraries in both C++ and Java

Reasons Against:

  • Question: Might using protobuf reduce the likelyhood of the CardSpace team collaborating with us and potentially adopting the CardSync Protocol? <-- an outcome that would increase interoperability in the industry

LICS design points

  • Need to have verification/authentication between selector and browser plug-in to LICS. This needs to consider factor in design of LICS.

Top Level Architecture

As you can see we have formalized the separation of presentation from core services:

Unified-selector-1.1.116.png

Notes:

  • We introduce the notion of a "Component Set" -- a set of components
  • This architecture would run on Windows, Mac OSX, Linux and (with further work) potentially smart phones
  • The "Selector UI" component would be either GTK, Cocoa or AIR-based, but the underlying Local I-Card Service would be common.

Phase 1

The first objective is to perfectly align the existing Components with the above diagram.

  1. Jeesmon: Split the shared tcpserver project into multiple projects to align with above. Suggestions for new names:
  2. Jeesmon: Merge the currently separate HSS connectors into .higgins.hss per the following ticket 258504
  3. Jeesmon: Split the AIR Selector code (org.eclipse.higgins.air ) into two project files
    • org.eclipse.higgins.selector.ui.air - selector UI in AIR/Flex
    • org.eclipse.higgins.selector.client.air (will eventually be replaced with a common .higgins.lics Local I-Card Service in C++) - selector services in AIR/Flex
  4. Andy: Split GTK/Cocoa Selector component into smaller pieces. Here's the first split:
    • Leave "org.eclipse.higgins.cbselector" project as-is (for Higgins 1.0 use)
    • Copy just the GTK-based user interface portion of .cbselector (shown in a box here) into a new project .higgins.selector.ui.gtk as the first alternative implementation project within the new Selector UI component shown above.
    • Copy the rest of the .cbselector project into a new .higgins.lics (Local I-Card Service) component
  5. Change GTK-based Selector to use standard Higgins HBX

Phase 2

Local I-Card Service

The client side of phase 2 involves creating the .higgins.lics project by copying the "Identity Selector Service" from the .cbselector and replacing the its i-card store with an i-card cache

Overview:

Lics-1.1.122.png

Synchronizing Card Store (Component Set)

One possible way to implement the I-Card Cache:

Synchronizing-card-store-1.1.118.png

Server Modifications

A new, RESTful binding I-Card Sync Web App component would be layered over an enhanced I-Card Service.

Server-mods-v4.png

Card Sync Protocol

Phase 3

This phase is about adapting the AIR Selector to the new architecture.

Back to the top