Breaking url-bar search

Gavin Sharp gavin at
Mon Aug 19 17:48:32 UTC 2013

On Mon, Aug 19, 2013 at 8:42 AM, William Pietri <william at> wrote:
> Ok. Yes, sorry for being imprecise, I meant "no direct data on the use of
> this feature". Or, perhaps to be more neutral with respect to what's a
> feature, I should say "no direct data on how many users would perceive this
> as a breaking change".

Yes, that's fair.

> Ok. That back-of-the-envelope calculation was my way of trying to reconcile
> your theory that there are only thousands of users affected with the fact
> that within a couple of weeks 760 people have found and installed this
> extension. That seems an impossibly large uptake rate to me given the number
> of steps involved and typical user behaviors. How would you reconcile the
> two facts?

I'm not sure there's much use in speculating, but here goes. The two
add-ons available to address this problem
( and are
well-linked from support articles describing the change, blog posts,
etc. They are easily findable for affected users doing a Google search
for "keyword.URL". So it makes sense that they will have been
installed by many such users. If 20000 users used keyword.URL or cared
about having distinct engines for the search bar vs. the URL bar, it
seems possible that ~2000 of them would find and install one of the
add-ons that restores the functionality.

> Also, my theory is that the number of people negatively impacted by this
> change is people who:
> use the url bar to search Google,
> have changed the search box to something else, and
> continued to use the url bar for Google.
> I believe that is going to be much more than 0.002% of your user base. Do
> you?

Perhaps, but again I'm not sure I see much value in speculating in the
abstract. We can argue about specific made up numbers, but what is the
end goal? Should we revert course on the changes we made? That can't
happen at this point. Should we attempt to address the use case in
some other way? I would welcome your suggestions, but I'm skeptical
that it's a use case worth investing a lot of effort into addressing.
Should we attempt to actually measure the impacted users? I'm not sure
how we would go about doing that, and it seems like there are likely
more pressing things to measure. But I welcome your thoughts on
proposed steps forward from here!


More information about the firefox-dev mailing list