Day 4: Lawyers are from Mars, Engineers are from Venus

This is the coverage of the remedies trial in US v. Google. I’ll be in Virginia writing every day on the courtroom activities.

Day 4: Lawyers are from Mars, Engineers are from Venus

They say being in the army is 90% boredom and 10% terror. Not too different from antitrust court today. Let’s do boredom first.

The continued testimony and cross examination of yesterday’s witness, software expert Professor John Weissman, took up 6 hours of my life that I’ll never get back.

Professor Weissman seems eminently qualified to evaluate the possibility of a software spin-out and open sourcing options. With the enormous qualifier that he’s an outsider looking in at the complex borg of Google. As a result, an aggressive, skilled attorney, like those Google has on (no doubt an enormous) retainer, can spend hours poking holes and carving exceptions to his opinions.

Did Professor Weissman evaluate every possible dependency of the AdX code on other Google products? No. Can Dr. Weissman assure us that the migration will work no matter which cloud the (at this time unknown) buyer chooses? No, of course not. Asking him to do those things is impossible, as any engineer would admit.

The net effect was that Google scored a bunch of points in the courts’ mind about the uncertainty and risk inherent in the DOJ’s proposals.

Also we had this gem of an exchange when a Google lawyer was reading a document that indicated C# as a programming language:

Lawyer 1: I think this says “C…ampersand”

Lawyer 2: No, I think its “C..hashtag”

There was also an instance in which that same lawyer pulled up an article from Vulture that claimed “Cobol is easy to learn and understand.” I swear I checked that the article’s date was not April 1st, because otherwise I don’t know what world I live in.

Witness #10: Tim Craycroft

So we spent all morning learning that spinning AdX was impossible, and that open sourcing the “Final Auction Logic” of DFP was also impossible. Totally impossible! Can’t be done! We built the pyramids, but AdX spin? Impossible.

Then Tim Craycroft, the VP in charge of display products at Google, took the stand.

Turns out Google has been evaluating doing pretty much all the things the DOJ has been asking for and they are all imminently possible. In fact, Google might have done them on its own if not for all these pesky lawyers getting in the way.

Project Sunday: Sometime in 2020 Google did an evaluation of an AdX and DFP spin-out and found that it was feasible. The document describing this project was not available to the court, as (I believe) it was work product related to possible settlement talks with the EU, and thus was ruled inadmissible.

Project Monday: Following Project Sunday, in 2021 Google created Project Monday, which contemplated fully shuttering AdX, and transforming DFP into a Google Cloud product that publishers could license. Fan me off, cause I’m getting the vapors.

In just Tuesday’s newsletter I wrote about the options for a spin-out and included what i called “Option 3: Consolidate and/or Shut AdX.” Turns out, Google was on board with this plan! You might think this is crazy, but roughly 60% of AdX is just a wrapper on top of Google Ads, and another big chunk is DV360. Build a pre-bid adapter for those products and you probably cut no more than a couple hundred million in revenue.

Regarding the DFP-in-GCP idea, I kind of like it for big pubs. And if I was a betting man, I’d guess that Google would still require use of their tags so their on-page data advantage would remain. Pretty sneaky, sis.

2023/2024 Analysis: Nothing came of Projects Sunday or Monday, but Tim testified there was another analysis of options that, interestingly, concluded just in advance of last Fall’s DOJ trial. In this analysis, they looked at an AdX sale, open sourcing of the DFP logic, and having DFP bid into prebid — almost eerily similar options to what the DOJ is currently asking for a year later.

AdX Business Transfer: In this 2024 analysis they contemplated a “Business Transfer” of AdX to another buyer. In this scenario, no source code would be sold, but “reference code” and technical assistance would be given to a buyer to reproduce AdX on their own infra. The customers and business relationships would be migrated and in the end Google would shut down their AdX. They estimated this process would take a maximum of 2 years to start the process, and up to 2 years to migrate all the customers.

Open Sourcing DFP: Google contemplated taking the auction logic of DFP, open sourcing it, and then letting publishers host custom versions themselves which the DFP service would call out to in real time. They estimated that including “industry adoption” this might take up to four years. Ed note: This is soooo cooooool. They should totally do this.

DFP Bidding into Prebid: This is almost certainly going to happen under all scenarios. They spec’ed it a year ago and thought it could take 2 years to do. But in the meantime nothing has happened.

Are we getting close to a settlement?

Although a Google employee, Tim was called as a witness for the DOJ, so the cross-exam was done by the more friendly Google team.

Tim was asked to review the Google proposed remedies (good, honest, solid) versus the DOJ proposed remedies (bad, impure, vague). What struck me, however, was that just as the lawyers had poked holes in all of the engineering evaluations earlier in the day, Tim’s testimony was the exact opposite — it was an engineer poking holes in the DOJ’s legal docs. Neither of these analyses are in any way helpful to getting to an acceptable solution!

But there was actual progress. I don’t know if this is normal, but when Tim was asked about flaws in both proposals, he essentially started negotiating on the stand. Here’s some things I heard:

• If having PMPs and PG deals bid into prebid was important to publishers, he would be open to that.

• If restricting AdWords from bidding directly into DFP was important, he was open to that.

• Same with DV360, though he wanted to reserve the right to bid into prebid.

• 3 years for monitoring conduct (Google’s proposal) was “not a magic number for us”

• The open source auction definition was too broad and non-specific.

• Regarding behavioral remedies that Tim had many criticisms of as written, he said “we’re very open to implementing clarified versions of these.”

These parties are not that far apart. Obviously, the spin of AdX is the big remaining issue, but we already know that Google is willing to do it! They wouldn’t have studied shutting it down entirely if they were vehemently against it, would they?

Another vote for shutting it down

Judge Brinkema never ceases to amaze or surprise me. During Tim’s testimony this idea of fully shutting down AdX came up again and she said:

Honorable mentions

• “Cobol is easy to learn and understand”

• Borg vs Kubernetes comparison