Conversation
This changes PipelineId to be a UUID and removes the concept of SubpageId. Iframes are now able to generate their own pipeline ID during navigation which simplifies a lot of logic in other parts of the code. Other parts of the code, such as the compositor and layout only ever deal with pipeline ids.
highfive
added
the
S-awaiting-review
|
|
|
@jdm @Ms2ger What are your thoughts on this change? I thought I'd prototype this and I think it worked out OK. All the ref, wpt and css tests pass, and real web sites I've tested seem to work OK. I haven't been able to test with servo-shell yet, as it's broken in other ways at the moment. But I wanted to get your thoughts on whether this was a reasonable approach, and if so I'll do a bit more testing before I get you to review it. |
|
I admit that this change makes me nervous due to the possibility of UUID collision, but I don't actually know how concerned I should be about that. |
|
☔ The latest upstream changes (presumably #7423) made this pull request unmergeable. Please resolve the merge conflicts. |
highfive
added
S-needs-rebase
|
The last two paragraphs of https://en.wikipedia.org/wiki/Universally_unique_identifier#Random_UUID_probability_of_duplicates are informative. |
|
True! That being said, I am also concerned about multiprocess Servo where each content process is using a different seed, which the article does call out as a separate problem. |
…eate pipeline IDs. This is currently the constellation, and each script thread that is created. This guarantees that pipeline IDs are unique, without relying on UUIDs. Each time a new script thread is created, the constellation passes through a new namespace identifier.
|
@jdm What about something like the last commit? PipelineId becomes a tuple of (namespace id, index) and there is a namespace per script thread (plus the constellation). I'm not sure about the naming of the structs, but the concept seems reasonable to me. Thoughts? |
|
I don't have a good mental model of what these are used for. Could an attacker gain anything by causing collisions? |
|
I'm not sure - perhaps - but I prefer the changes in the last commit anyway, which remove the reliance on UUIDs and guarantee a unique pipeline ID. |
|
Removing in favour of doing this in small increments - first step is #7807 |
6 participants
This changes PipelineId to be a UUID and removes the concept of SubpageId.
Iframes are now able to generate their own pipeline ID during navigation which simplifies a lot of logic in other parts of the code.
Other parts of the code, such as the compositor and layout only ever deal with pipeline ids.