Bug 1769585 Comment 1 Edit History

Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.

3. is a non-problem imo, because if the user types a single letter and picks a url, that's a sign they may want that.

Doing 1. it won't be possible to train the system to autofill a url just typing a single letter or couple of them.

On the other side, if we do 2. we may match less often at the beginning, because the user has to type the exact word they typed in the past before the behavior kicks in. Though, the user can then train the system by typing the same 1 or 2 letters and picking the same url.  The problem could arise from signal persistence, we retain adaptive history for 90 days of use. That long time may complicate re-training the behavior. I suspect 30 days would work better for this feature: that'd mean changing the threshold from 0.1 to 0.47.

My best guess would be to go for 2 and also rise the threshold to make training easier.
We could also introduce a minChars pref that by default would be 0 (disabled) but if we rise it to a positive value the user has to type the whole word but no matchng below the min chars required. This may allow for an alternate testing path.

How does this sound?
Point 3. is a non-problem imo, because if the user types a single letter and picks a url, that's a sign they may want that.

Doing 1. it won't be possible to train the system to autofill a url just typing a single letter or couple of them.

On the other side, if we do 2. we may match less often at the beginning, because the user has to type the exact word they typed in the past before the behavior kicks in. Though, the user can then train the system by typing the same 1 or 2 letters and picking the same url.  The problem could arise from signal persistence, we retain adaptive history for 90 days of use. That long time may complicate re-training the behavior. I suspect 30 days would work better for this feature: that'd mean changing the threshold from 0.1 to 0.47.

My best guess would be to go for 2 and also rise the threshold to make training easier.
We could also introduce a minChars pref that by default would be 0 (disabled) but if we rise it to a positive value the user has to type the whole word but no matchng below the min chars required. This may allow for an alternate testing path.

How does this sound?
Point 3. is a non-problem imo, because if the user types a single letter and picks a url, that's a sign they may want that.

Doing 1. it won't be possible to train the system to autofill a url just typing a single letter or couple of them.

On the other side, if we do 2. we may match less often at the beginning, because the user has to type the exact word they typed in the past before the behavior kicks in. Though, the user can then train the system by typing the same 1 or 2 letters and picking the same url.  The problem could arise from signal persistence, we retain adaptive history for 90 days of use. That long time may complicate re-training the behavior. I suspect 30 days would work better for this feature: that'd mean changing the threshold from 0.1 to 0.47.

My best guess would be to go for 2 and also rise the threshold to make training easier.
We could also introduce a minChars pref that by default would be 0 (disabled) but if we rise it to a positive value the user has to type the whole word but no matching below the min chars required. This may allow for an alternate testing path.

How does this sound?

Back to Bug 1769585 Comment 1