Runrig

Appearance

Appendix: Architecture ​

DRAFT NOTE

This page is still in a very early stage of draft, mostly scattered notes, links and references.

Provisional outline:

  • Data Provider
  • Service Platforms
  • Local-fist & Self-hosted Applications

System Characteristics ​

DRAFT NOTE

The main points here warrant further evaluation and debate, to be sure, but eventually it would be nice to represent the agreed-upon system characteristics in a tabular format, preferably somewhere at the top of this appendix, while the details and citations should be integrated into the body text, across the relevant page sections.

Data Provider ​

Storage Provisioning ​

It may also be helpful to structure storage costs in a way that a platform or service on a platform can provision a block of storage for multiple users on a Solid server, so they can build that cost into their own fees and/or shared costs.

This is actually quite similar to how the private Solid provider, use.id, designs their Provisioning API:[1] [2]

http
# @name provisioning
#
# Registers a new WebID, if it does not already exist, and creates a new WebID profile for it.
# Value for Slug is used for creating the WebID -- here: https://dev.use.id/tom
#
# Responds with 201 Created, a Location header to the new WebID, and the contents of the WebID profile as a body.
# Responds with 409 Conflict if the WebID already exists.
#
POST /provision HTTP/1.1
Host: https://dev.use.id
Authorization: DPoP <api-token>
DPoP: <dpop-proof>
Content-Type: application/json
Slug: tom

{
    "email": "tom-stevens@email.com",
    "password": "passwd1234"
}

other considerations...

  • one person, one pod?
  • use of homomorphic encryption and synthetic data?
  • auth flows

Why Solid? ​

As described in Solid Project's documentation:

One of the core ideas behind Solid is to make data independent from applications, so that one can be in control of his/her own data and share it with the apps of his/her choice. For this to be possible, the same piece of data must be understood consistently from one app to another. This requires agreeing on the meaning of the data, as conveyed by its description. Therefore, to make data reusable, it should be described with vocabularies that are widely used and known.[3]

Service Platforms ​

Distributed, Federated Services & Applications ​

When it comes to hosting services and applications, I think it makes sense to host and operate that software on a smaller, regional scale and in a more distributed manner, compared to how data for that software will be hosted and stored. This is the second of the "primary responsibilities" listed above.

Mostly, this will be done to allow greater localized control of the actual applications and services they host, which may require additional storage for more sensitive information like billing, payments and personnel data. They should also be empowered to modify and extend their software as befits their particular needs. Likewise, they should be granted a high level of local autonomy in determining how costs are shared, how prices are set, and how the overall governance of their regional platform is administered. However, they will still benefit from services and resources that are shared nationally and globally, such as Solid pod storage. By the same token, they will also be obliged to contribute finances or resources globally and abide by the terms set by those broader governing bodies, though ideally any restrictions would be kept minimal and enforcement rarely called for in practice.

It might be helpful to illustrate this with a stack diagram, like the ones used for illustrating architecture of the Linux kernel and similar technology systems:

Stack diagram of the Linux kernel

References ​

  1. use.id. "use.id for Business". ↩︎

  2. Van den Briel, Abel. use.id Forum. "Design of the Provisioning API - Feedback welcome" (login required). Jan 31, 2022. Further in that forum post it states that "<api-token> is an access token, retrieved from the client credentials flow" and that "<dpop-proof> is unique per request". Another commenter, identified as use.id staff, cautions that sending passwords is not intended for production use and is only provided as a temporary example while passwordless flow is still being implemented, but as of April 2022, that system was not ready for testing. ↩︎

  3. Solid Project. "Vocabularies overview". ↩︎