@botiverse/porthole 0.15.1 → 0.15.3

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md CHANGED
@@ -4,6 +4,48 @@
4
4
  section); the reasons for every change are in [docs/DECISIONS.md](docs/DECISIONS.md), by the
5
5
  P-numbers below.
6
6
 
7
+ ## 0.15.3
8
+
9
+ - **A nested page's real presses count, each of them** (P148, amends P111). A nested page's shell
10
+ knew of a press in the page only as its own activation turning on, and an activation stays on
11
+ for 5 s: a second press in that time, or a press within 5 s of one on the host's page (a
12
+ dialog's Cancel), turned nothing on, and `comments.open` from it was refused ("This needs a
13
+ click"), or a move into the page taken back. The page's runtime now tells its shell each press
14
+ the browser itself made (a mouse down; a finger or pen up; Enter or Space, not held, on a native
15
+ control that isn't editable; a click on one), over a `MessagePort` the shell sends at each load
16
+ and the runtime takes before any content runs. The shell says when it heard the last
17
+ (`porthole/shell/state`: `pressed`, `presses`), and the host takes it as it takes the
18
+ activation turning on: for a second, with no press or key on the host's page since, and under
19
+ P120's and P122's rules. Nothing a page dispatches is a press, no click while a key is down is
20
+ one, and no content holds the port;
21
+ where the runtime can't install all of it, nothing is told and P111's rule stands alone.
22
+ Only for nested pages (`nested: true`): a served page that isn't nested is as before, and in
23
+ full Chromium ADOPTING's advice for it stands (nest it).
24
+ **For a host:** no API change. The nested shell's script and the shell runtime every served
25
+ document inlines change (the nested shell's `script-src` hash with it; `nestedShellResponse`
26
+ and `/server` compute it, as before). Both sides need 0.15.3 for the change to show: the host's
27
+ page (`createHost`) and whatever serves the content origins. Either side alone behaves as
28
+ 0.15.2.
29
+ - **The helpers that take host origins check them** (amends P145, after the audit of 0.15.1).
30
+ `statusPage`, `doorResponse`, `doorStatusResponse` and `nestedShellCsp` took `hostOrigins`
31
+ unchecked: `:0` passed, and a string with a `;` went into the CSP header as written. They are a
32
+ host's own configuration, never content's; each helper now throws a `TypeError` on one
33
+ `hostOriginProblem` refuses, as `createServer` and the host kit already did. `isHostOrigin` now
34
+ refuses what `hostOriginProblem` does (it took `:0`, a port past 65535 and a host the URL parser
35
+ rewrites). **For a host:** a content Worker of your own that passed a malformed origin to one of
36
+ these now gets an error where it got a header no browser would match.
37
+
38
+ ## 0.15.2
39
+
40
+ - **Marks ride the document's scroll** (P147, amends P104 and P132). A commented passage's
41
+ highlight and its marker were placed in the viewport and put right again from the scroll event:
42
+ in the frames a browser scrolls before its main thread runs that event, they trailed the text.
43
+ They are now placed down the document and moved by a scroll-driven animation
44
+ (`animation-timeline: scroll(root y)`), which the engine runs with the scroll itself. Where there
45
+ is no scroll timeline (Firefox today), in a scroller of the page's own, and sideways, a mark
46
+ follows from the scroll event as before. A mark on a `position: fixed` or stuck `sticky` element
47
+ now trails for those frames instead (accepted: P147, decision 6). Nothing for a host to change.
48
+
7
49
  ## 0.15.1
8
50
 
9
51
  After Ferryman's audit of 0.15.0 (P145, amends P143 and P144): no P0, no escape, no new egress,