Skip to content

Tried, and not yet

What has run against a real workspace, and what has not. This page is the one record of it. Everything else Caland does is held by its tests: against a stand-in for the workspace, with the server asked over real HTTP and the page driven in Chrome.

The runs below were made on 2026-10-06, on one test workspace, with the page as it was that day, between 0.5.0 and 0.6.0. No run of a later release on a workspace is on record.

2026-10-06: reading

Caland connected with a profile and was ready in 2.3 seconds; the page was painted after 60 ms. That workspace had no secret scopes then, so what was shown was the empty page.

2026-10-06: changing

The real page, driven in Chrome, in a throwaway scope it made and removed. Each step was checked through the SDK, not through Caland. The workspace had the scopes afterwards that it had before.

What Held?
A scope is made, and deleted with what was in it yes
Text is stored as it was typed yes
A file of every byte there is (NULs, a byte order mark, CRLF, random bytes) is stored byte for byte and shown as what it is yes
A certificate is described, and stored as the file was yes
A rename, and the original put back yes
A delete, and the secret put back byte for byte yes
A grant given and taken yes
A rename that only changes the case is refused, and the secret is still there yes
Case-Key and then case-key is one secret, and a scope asked for in capitals "already exists" yes

20 checks, all held, and run again after the fixes of that day's review.

2026-10-06: which workspace

Plain caland opened the page on the test workspace. W listed the one profile the machine has, found in ~/.databrickscfg; chosen again, it connected again.

Not yet

  • A Key Vault-backed scope. The workspace has none.
  • A workspace where the person may not write.
  • The choice between several workspaces, and signing in by an address. That opens a browser to sign in with, which nothing here can drive.
  • .env in and out, the stale list and the two lists of access, on a real workspace.
  • Firefox and Safari. The page is driven in Chrome only.
  • A real drag and drop of a file onto the form.
  • Windows.
  • Stopping by itself while a real sign-in in the browser waits. The tests hold it with a sign-in that never finishes, made by a stand-in. The SDK's own wait was not run.
  • A real DNS name pointed at the machine, and a Mac whose .html files open in something other than the browser. There caland --no-open gives the link.