Open
Bug 1911171
Opened 2 months ago
Save ML Model Cache in a dedicated principal
Categories
(Core :: Machine Learning, enhancement)
Core
Machine Learning
Tracking
()
NEW
People
(Reporter: atossou, Unassigned)
References
Details
Currently, the IndexedDB cache is stored in the system principal "chrome".
I propose saving it into a dedicated principal, such as "about:inference".
This change offers several benefits:
1) More accurate estimation of disk space usage by the ML modules.
2) More granular control, allowing us to assign lower priority to that principal's storage so the browser can remove it under pressure conditions.
This bug was separated from https://bugzilla.mozilla.org/show_bug.cgi?id=1910116 to discuss an appropriate origin. Using https://model-hub.mozilla.org/ is another possibility, but "about:inference" seems more suitable.
To solve this bug, the main change is to replace DEFAULT_PRINCIPAL_ORIGIN in Modelhub with the chosen value.
You need to log in
before you can comment on or make changes to this bug.
Description
•