0016 Link parameters in the page's query string¶
Sources: #619, #615, eizedev/powershell-universal-components#144
Context and Problem Statement¶
A link names its data in the fragment: src.id, and src.params as far as the page lets them
through with allowParams (sharing). The
browser's Copy link writes them. create_link writes the source id only and takes no parameters.
A host application that fetches rows on its server side (0015) never sees the fragment. The first report takes a parameter, such as a customer. Where does a link carry it?
Considered Options¶
- In the fragment's
src.params, and the page sends them to its backend. - In the page's query string.
Decision Outcome¶
Chosen option: the query string, because a host application's backend reads it without code in the browser.
create_linktakesparamsand writes them into the query string. The view stays in the fragment.- A link carries only the parameters the operator lists for the source (
sources.<id>.linkParams). A parameter bound to a sign-in claim never goes into a link. - The server checks the values with the source's parameter rules when it makes the link. The rows route checks them again with the rights of the person who opens the page.
- The list binds only the links the server makes. Anyone can write a URL by hand. So
GET /sourcesreturns the list, and the page passes on only those names. - The fragment's
src.paramsstay as they are, for what a browser reads, such as a workbook'ssheet. Where both name a parameter, the query string wins.
Consequences¶
- Good: a host application reads the parameter on its server side.
- Bad: values in a query string land in access logs, browser history and the
Refererheader. List only parameters whose values may show up there. The host page sends a strictReferrer-Policy. - Bad: a parameter name can collide with the host page's own query. The operator chooses the names.
- Bad: in stage 1 only
create_linkwrites the query string. The browser's Copy link drops it unless the host gives the report a share address that keeps it.