Skip to content

Latest commit

 

History

History
52 lines (37 loc) · 4.44 KB

README.md

File metadata and controls

52 lines (37 loc) · 4.44 KB
Raw

Open Network HTTP COPY Symlink Module for FS Store

The intention of this module is to provide support for the COPY command to help achieve compatibility to the Solid project.

To allow for clean user experience, we sometimes require very quick operations to allow for instant user interaction, one type of application implementation that will be seen built on top of Solid is to define state machines by utilising the powerful graphing ability of RDF and FaaS.

To achieve a quick response time we could define a way to offload COPY requests to a backing service, but we will also need to be able to make the resources available from our service immediately if requested.

There are different levels of acceptable outcomes for a COPY request, here's a quick list of use cases we will cover:

  • Agent makes a COPY request with header Source a resource URI, agent requires the resource to be cloned immediately, and for a shallow copy to be made (this means they do not want us to follow linked resources at all), service is expected to return status code 201, with the Location header the URI of the cloned resource.
  • Agent makes a COPY request with header Source a resource URI, agent requires the resource to be available immediately, and for a full copy to be made for the origin (this means they do want us to follow linked resources from the same origin), service is expected to return status code 201, with the Location header the URI of the symbolic resource. Service must follow an agent accessible pre-defined cloning scheme.

Above, we have defined an entry point for a new set of rules, what a cloning scheme is. We want to seperate this concern as agents that have different requirements will require different ways to clone information.

Example Copy Scheme

COPY:

  1. Service will make a HEAD request to Source
  2. If the HEAD status code is not between 200 and 299, and 301 to 308, then return 400 with header Warning: Resource not available - ${headResponse.status}
  3. Service will make a COPY request to backing service for Source
    1. Backing service will create new random Pointer URI
    2. Backing service will return status 201 with header Location: ${Pointer URI}
    3. Backing service will set resource at ${Pointer URI}.meta to Downloading
    4. Backing service will make a GET request to Source
    5. If GET return status is not between 200 and 299, and 301 to 308, then set resource at ${Pointer URI}.meta to Warning: ${getResponse.text()} and abort.
    6. Backing service will store download resource at Pointer URI once downloaded
    7. Backing service will set resource at ${Pointer URI}.meta to Downloaded
  4. Create resource at ${Target URI}.meta describing the location of the new resource at Pointer URI
  5. Return 201 with header Location: ${Target URI}.

GET:

  1. Service will check if ${Target URI}.meta describes the location of a resource, if not continue as normal
  2. Service will make a GET request to ${Pointer URI}.meta to see the status of the resource
  3. If GET response is Downloading or Downloaded, then return status 302 with Location: ${Pointer URI}
  4. If GET response is Warning, then return status 404 with header Warning: ${getResponse.header.get("Warning")
    1. Backing service will schedule ${Pointer URI}.meta to be deleted in n days
  5. If GET response is not Downloaded, return
  6. Service will make a COPY request to ${Target URI} with header Source: ${Pointer URI}, service requires the resource to be cloned immediately
  7. Service will delete the location described in ${Target URI}.meta

Why would this be useful?

  • Cloning sets of information from a remote service to a local trusted space, for example if you wanted to browse the internet offline and already know what set of information you'll be using.
  • Cloning state machines defined in RDF graphs
  • Cloning resources defined within a graph, this could be a full website definiton with backing state machines or defined dependent resources (for example by way of a Ansible or Pupet, or a set of hyperlinked HTML documents
  • Cloning documents between personal storage services
  • Providing a common strategy whether the primary service is local to the agent, or on remote infrastructure