Stop hard-failing on unknown imp.ext values

[bretg]

One of the RTD modules starting putting data on imp.ext, and PBS is choking on it.

This isn't right... imp.ext should be for everyone to use... PBS shouldn't assume that every value in that ext must be a bidder to the point where it hardfails.

Instead, if an unknown value in imp.ext is found, PBS should do the following:

1. If imp.ext.prebid.bidder exists, then do nothing. The requestor has moved to the new syntax where bidders are in imp.ext.prebid and there's no need to check imp.ext anymore. Let it through.
2. If imp.ext.prebid.bidder doesn't exist, then it's ok to warn about unknown values, but the request should otherwise proceed.

Note: we already make exceptions for many imp.ext values: ae, all, context, data, general, gpid, prebid, skadn, and tid. We can keep these special cases to reduce the number of warnings emitted.

Test cases:

1. Add imp.ext.foo and imp.ext.prebid.bidder.appnexus to the request. There should be no warnings. imp.ext.foo should be passed to bidders.
2. Add imp.ext.foo and imp.ext.appnexus to the request. There should be a warning about an unknown bidder "foo". imp.ext.foo should be passed to bidders.
3. Add imp.ext.gpid and imp.ext.appnexus to the request. There should be no warnings. imp.ext.gpid should be passed to bidders.

[Slind14]

One option might be adding it to the list of reserved keys but that is an inclusive list:

https://github.com/prebid/prebid-server/blob/7702a78118cf56ebf2a4e998f4757b01640ef404/openrtb_ext/bidders.go#L251

[bretg]

adding it to the list of reserved keys

I suppose that would be faster, but I'd like to fix this once and for all

[bretg]

This is a high priority because imp.ext.igs is now a thing.

[vraybaud]

Hey, gentle bump Please do we have any update about this issue ? We want to use Fledge with PBS but it looks like it's not possible with latest version of paapi PBJS module because of imp.ext.ae being replaced by imp.ext.igs.ae.

[bretg]

Should be in PBS-Java this week. @bsardo - is this in the queue for PBS-Go?

[bretg]

Released with PBS-Java 3.9. Discussed in committee - this is on the radar for PBS-Go.

[bmjoju]

@bretg What is status on this? We now have to remove the paapi module, so we can update to version 9.

[bretg]

I pinged the PBS-Go team with a reminder. It might get done sooner if a community member was willing to port it to Go from PBS-Java...

[bsardo]

There's already a PR open for PBS-Go. #3878 addresses this. We're actively reviewing it.

[bretg]

Released with PBS-Go 3.3