727085 - Stop breaking history in search results/kb

Status: RESOLVED FIXED

(support.mozilla.org :: Search, defect)

Description

[James Socol [:jsocol, :james]]

Right now we push a bunch of things into the query string of search results for metrics, with the effect of making results nearly always appear unvisited. From bug 719981 comment 16:

> This is stuff like, e.g.: "?s=crash&r=0&e=es&as=s" which means:
> * searched for <crash>
> * this article was the first result
> * the search engine was elastic search
> * and the user found the article via search
> 
> (as=s could probably now be inferred from the other info but that doesn't help
> with this problem at all, and it's convenient for compare with other sources
> like as=u.)
> 
> I have no idea how to fix this except maybe we can trash some of those metrics,
> since I don't know that we're actually using any of them right now, and a lot
> could probably be gathered from logs and referrers.
> 
> It might be good, if we do that, to also do _something_ with as=u, like get
> rid of it and use referrers with matching #^/1/# to count those, so articles
> visited from the UI also show up as 'visited' in history.

Before we remove things, we need to understand which of those metrics, if any, we're actually using, and which we can get rid of. If we're using something we'll have to find a new way to get it.

Comment 1

[Ricky Rosario [:rrosario, :r1cky]]

We can do what google seems to do. The href points to the destination page but there is a javascript click handler that sends the request to a redirector. We would have to depend on log data for metrics if we wanted to do this. Webtrends would never see the query string values.

Comment 2

[James Socol [:jsocol, :james]]

If we did that, we could have the redirector record some bits, like average result rank, instead of getting them from logs.

On the other hand, we could already get a bunch of data from referer values in logs.

But if we're not actually using any of these, then meh.

Comment 3

[Will Kahn-Greene [:willkg] ET needinfo? me]

I've removed the e= one when I ripped out all the Sphinx stuff.

Cww: That probably affects your metrics. All search results and CTR things and links are now ES links. There's no more e=es/e=sph in the urls anymore.

Comment 4

[Michael Cooper [:mythmon]]

Our search urls are clean now, and links turn purple.