Disabling AutoComplete

Consideration should be made for the user who does not want to see autocomplete results below the address bar. This is a relatively common request for users migrating from the Suite, where this is an option (a pref: browser.urlbar.autocomplete.enabled). Currently a rather ugly hack is required to accomplish this -- and the code matching against history still runs, just with no visible effect. If performance is already an issue taken into account when querying the history, this seems like a simple thing to check for first. --Unarmed 11:17, 4 Jan 2006 (PST)

Personally, I can't imagine this will ever happen.

-- Brett Wilson 16 Aug 2006

Search bar suggestions behaviour

I think that the suggested links should not just be those where the domain name starts with the phrase that’s being typed, like is the case currently. Instead it should show any link that contains the phrase in part of its information set (including page title and url), sorted by a combination of ‘popularity’ (a combination of number of visits and how long ago those visits were), and what part of the information contains the phrase (e.g. if it’s in the domain name then it gets a slightly higher weight). This will give much better suggestions that cover more than just the domain name, and also let the user to find short phrases that identify a site more uniquely.

To illustrate, if I currently type ‘places’, when I get as far as ‘pla’ I only get a few suggestions from ‘player.omroep.nl’ (which I relatively rarely visit, and is a player pop-up too) and ‘playstationhomeinsider.googlepages.com’ (which I’ve probably not visited more than once), and once I fully typed ‘places’ there are no suggestions at all. Even though I just visited these very relevant pages with Places in both the title and the URL, and have read various blogs on the subject in the past. Grauw 23:41, 13 May 2007 (PDT)

Page information stored

In addition to the page title and URL, you should also store the contents of the ‘description’ and ‘keywords’ META tags contained in a page, and use it for the search bar suggestions. This so that if a user types ‘html’, www.w3.org will appear in the suggestions, or if ‘ipod’ is typed, apple.com would show up.

Having this would put the descriptions and keywords on many pages to good use, better expose them to the user (and thus giving them a higher probability to be properly up-to-date), and also provide incentive for web site owners to add these. It will also discourage the page author from abusing the page title to provide a description and keywords for the page by giving them a better means to do so (see e.g. [1]).

Ideally, you would also extract keywords from the page content, by e.g. doing a frequency analysis of words, removing certain common words like ‘the’ and ‘think’, and storing the top-100. But I think that that’s kind of complex, and for now storing just the meta-information (in addition to the page title and URL) will suffice. Grauw 23:41, 13 May 2007 (PDT)

Extensibility

The system should also allow extensions to provide additional data for the information set that the suggestions search through, like e.g. tags (for an extension that lets you tag pages, or one that can get tags from some microformat or whatever). They should also have the ability to specify a relative weight of importance for that data (user-provided data would generally have more importance than auto-generated data), and also be able to affect the global weight of a bookmark (which for example could interact with an extension which lets you rate pages). Grauw 23:47, 13 May 2007 (PDT)