Closed
Bug 1251573
Opened 9 years ago
Closed 8 years ago
Consider involving seer for rel=prefetch
Categories
(Core :: Networking, defect)
Core
Networking
Tracking
()
RESOLVED
INVALID
People
(Reporter: dragana, Unassigned)
Details
(Whiteboard: [necko-backlog])
Currently prefetch is implemented in uriloader/prefetch/nsPrefetchService.cpp
Prefetches are started when page load finishes and in case another page load starts all running prefetches are killed. We could be smarter about this. Also deciding when to start a prefetch and which prefetch to start first could be better. I think involving seer for this would be necessary. (If I am wrong just close the bug :) )
There is a draft (https://w3c.github.io/resource-hints/ - not finish jet so not everything is properly define) that adds parameter pr - "probability that the resource will be use". Currently we do not have notion of probability.
I'm certainly not adverse to this, but in its current incarnation, the seer wouldn't really be useful for this - the seer acts (and has data) on pages that are explicitly being loaded, not on signals like "hey, you're currently on page x, so you might go to page y soon, so go ahead and prefetch that" (which is that link rel=prefetch seems to be more geared towards). So, we'd have to think more about how to make the seer fit in with that kind of prediction. I'll put this on the backlog for now, as I think the idea has merit, just not sure when we (I) would want to spend a lot of time rethinking the seer :)
Whiteboard: [necko-backlog]
| Reporter | ||
Updated•8 years ago
|
Status: NEW → RESOLVED
Closed: 8 years ago
Resolution: --- → INVALID
You need to log in
before you can comment on or make changes to this bug.
Description
•