Document the api layer's operational models, and fix two headers that had gone stale against the code
This commit is contained in:
@@ -4,6 +4,19 @@
|
||||
// Script and document reader. Returns the full text of one .sh or .md file inside
|
||||
// SCRIPTS_DIR — the source view behind the scheduler page's script viewer and the docs tab.
|
||||
//
|
||||
// OPERATIONAL MODEL
|
||||
// Validate, resolve, read. The id is checked against a character class and an extension list
|
||||
// before it is joined to SCRIPTS_DIR, so nothing reaches the filesystem that did not already
|
||||
// look like a repo-relative path.
|
||||
//
|
||||
// The whole file is returned in one response — no ranges, no pagination. These are scripts and
|
||||
// documents, not logs; the largest is a few hundred kilobytes, and a viewer that had to stitch
|
||||
// pages together would be more machinery than the thing it displays.
|
||||
//
|
||||
// Every failure is a JSON body with ok:false, never an HTTP error code. The scheduler's viewer
|
||||
// and the docs tab both parse the response before looking at anything else, so a 404 would
|
||||
// surface as a parse failure rather than as "that file is not there".
|
||||
//
|
||||
// DESIGN PRINCIPLES
|
||||
// Two extensions, one endpoint.
|
||||
// Scripts and their READMEs are read the same way because they are read for the same
|
||||
|
||||
Reference in New Issue
Block a user