414661 - bypass caching by AMO when no recommended results are found

Status: RESOLVED WONTFIX

(Toolkit :: Add-ons Manager, defect)

Description

[Dave Townsend [:mossop]]

At the moment if the first pull of recommended results returns nothing compatible we are lost and can't ever find any more since the results are cached.

This will add a random code to bypass the caching allowing us to pull more results and so increase the odds of finding compatible addons.

This would be good to hit b3 depending on how quick I can knock this out and get reviewed.

Comment 1

[Michael Morgan [:morgamic]]

I don't think this is a good idea.  I would prefer to cache the master list and randomize selection of a few on the client side.

The other option is to disable caching altogether, and I don't want to do this either.  I am already nervous about making decisions like adding the random token when it creates a lower cache hit-rate for the sake of UI logic.

How hard is it to randomly select the recommended elements from the master list before handing them off to the add-ons manager?  Also, why not just list all featured add-ons when the tab is loaded (or at least just load the ones needed to fill the pane).

Comment 2

[Dave Townsend [:mossop]]

Ok so the plan of adding a random token doesn't actually work to get around the caching anyway.

(In reply to comment #1)
> How hard is it to randomly select the recommended elements from the master list
> before handing them off to the add-ons manager?  Also, why not just list all
> featured add-ons when the tab is loaded (or at least just load the ones needed
> to fill the pane).

The problem we have is that the current API returns a number of recommended results, the number being chosen by the client. The client then filters that result set removing add-ons that are incompatible or already installed. The problem is that right now retrieving even 25 results often yields just 1 that is compatible. Larger numbers of results seem to take longer to retrieve results increases fairly linearly up to about 4 seconds for 50 results.

I think the only thing to do here is to bump the number we pull by default and take the problem that we often will get no compatible results and know that it will alleviate as more add-ons become compatible.

Comment 3

[Michael Morgan [:morgamic]]

Would it help to set up a staging instance so we can test this elsewhere?  We can create faux add-ons that are compatible with fx3 until we have a strong enough sample set from the real add-ons universe.  Any thoughts on that?

Comment 4

[Michael Morgan [:morgamic]]

Second thought, we could always take a snapshot of production and make everything compatible with 3.*.

Comment 5

[Dave Townsend [:mossop]]

I don't see how that helps us. The point is for b3 users to see recommended addons that they can install.

Comment 6

[Michael Morgan [:morgamic]]

Yep, you're right.  Besides, it's too late for that now.

Comment 7

[Dave Townsend [:mossop]]

Ok by agreement with UX we will just leave this.