Comment by vbezhenar | Hacker News Reader

Comment by vbezhenar | Hacker News Reader

I always wondered, why websites don't deliver ads integrated? Why Google Analystics can't use server-side data? I don't want my user to load any unnecessary JavaScripts, if all data is already available in nginx logs. Why should my user issue additional HTTP requests to Google Ads, when I can issue those requests at server-side (and they'll work faster, because server interconnectivity usually is very good) and deliver the single complete web page to user, with all ads included. Images might load faster from different hosts now, because browsers might restrict connection count to the single host, but in HTTP/2 era single host with single connection will work faster, so loading ad images from website server will be faster too.

The only difference is how easy to integrate ads to your site. Adding <script> tag is certainly easier than tweaking server-side logic or setting up nginx module. But it doesn't matter for big sites, yet they employ the same technique, making their sites unnecessary slow, especially for low-bandwidth users.

Google does not trust sites as many will fake their numbers to make more from Ads. If Google directly talks to each user's IP it's much easier for Google to tell that each page load was a real page load.

In your approach they are stuck trusting the webserver and it's far easier to add fake page loads, this is somewhat less of an issue with PPC, but there are still issues.

On top of that advertisers want to track each user which is far simpler when it's a separate HTTP request.

Not to mention the algorithms and business behind which ad to serve where and when are worthy of an entire field of study all by themselves.

Read the Advertising chapter in Mining Massive Datasets - http://www.mmds.org/ - and then figure out how you'd tackle those problems when you distribute them onto other people's servers.

> Google does not trust sites not to fake their numbers. If Google directly talks to each user IP it's much easier for Google to tell that each page load was a real page load.

If we're talking about Google Analytics, the user wants to view correct data in the first place.

> In your approach they are stuck trusting the webserver and it's far easier to add fake page loads, this is somewhat less of an issue with PPC, but there are still issues.

That's true.

> On top of that advertisers want to track a user which is far simpler when it's a separate HTTP request.

Webserver is supposed to send all necessary user's data to Google server (like IP, UA, Cookies) and Google might ask Webserver to set google-specific Cookie.

Cross-site tracking won't be possible, that's true. Probably some standard could be invented for that (e.g. Google signs global inter-site cookie to deliver to the user and browser should send this cookie to all Google-enabled sites). That'll make easier for user to disable tracking, if user wants that.

My point is that if Google doesn't want to improve their scheme, may be other advertising company will want to improve that scheme. And Google might realize that users will follow that company, that's what competition is about.

And probably Google itself is not the evil here. There are many other ad networks with huge banners and slow loading scripts.

Google Analytics had a self-hosted version, or more specifically Google Analytics was a self-hosted solution when it was Urchin before Google acquired it.

You could still get the self-hosted version for a price until Google killed it in 2012.

The reason Google doesn't want you to do this is simple: they want the data.

The Google Analytics Measurement Protocol allows developers to make HTTP requests to send raw user interaction data directly to Google Analytics servers. YMMV, I haven't tried it yet but want to hook it into our app asap.

https://developers.google.com/analytics/devguides/collection...

Reply on news.ycombinator.com