The Internet Keywords feature needs to be removed from Mozilla. There are many and varied reasons for this, some of which are as follows. * It is Netscape-branded content in the Mozilla tree (bug 14532). * It stops the Enter key from working as expected (bug 65911). * it places an undue burden on the user to remember which words are keywords, and to enter the keywords very carefully once they are recalled (bug 45719). In short, Internet Keywords materially worsens the user experience, and appears to exist solely to drive users to Web sites of Netscape and its partners -- even if the users didn't want to go to those sites at all. Therefore its existence is not a feature, but a bug.

Matthew, I disagree with this bug. In Beonex Communicator, I changed the Internet Keywords provider to Google's "I'm feeling lucky" function. I used it myself quite a few times. You are right that this feature has UI short comings, most notably that it doesn't work for single words. But it is IMO a very cool feature to search for the "authorative" site for a certain term. I can enter "Beonex Communicator" into the URLbar, hit enter and being fairly sure that I get to the right page, even if I don't remember the URL. I used it in a similar way to get to the "PSM build instructions". No that this usage is very similar to the keywords Mozilla supports natively with Bookmarks (I can enter "bug 12699", when I have an appropriate bookmark for bugzilla), just that here, I don't have to enter any prefix. If Netscape uses the Internet Keywords feature to implement "quote MSFT" or "webmail myaimid" queries, that's their thing. As for myself, personally, I find Mozilla's Internet Keywords feature connected with Google's I'm feeling lucky feature very useful and cool, eh, "enhancing my web browsing experience" ;-P.

I think Ben B's usage shows that this is a reasonable feature to retain in mozilla. If we are currently using netscape as the internet keyword provider, perhaps we should consider switching to google.

Ben, I think the problem with Internet Keywords feature is that in Mozilla Browser there is no UI element to customize it's behaviour. I would love to be able to modify it to do an Internet Search. Or to direct it to the "I'm Feeling Lucky" option in Google. But It is not possible. It always goes to Netscape servers to evaluate the query.

This kind of discussion really belongs in the newsgroup. I can see the utility; the problem is that the UI is *so bad* I think it should be a showstopper. timeless: Why would Google be any better than Netscape if it's still hard-coded? This ought to be configurable by the user.

Daniel Mario Vega, I agree. But this bug calls to *remove* the feature, which is something completely different, actually, it's the opposite direction.

mpt, please rephrase and refocus the summary and direction of this bug. I think it's a little irresponsible of you to introduce such an obviously inflammatory 'bug report'. It is at best, a UI:DF bug, the component for which you are of course the default development contact. type 'southwest airlines' -- immediately get to http://iflyswa.com, yeah that's a stop ship allright. It's so bad IE copied it and spent thousands marketing their 'Internet Keywords' idea. btw the way the logic that "it stops the Enter key from working as expected" is flawed. When one types two words into the url bar eg 'honda accord' and presses enter what is it we expect to happen? Pre-internet keywords the answer is 'nothing' - an error alert. Only since the advent of Internet Keywords has that behavior led to a search. I'm not arguing for/against netscape here, I couldn't care less. Use google, use whatever, match the user's chosen search provider pref(already a bug). I'm just arguing that we don't need to spark inflammatory discussions in a bug report about removing a great feature that we just happened to get pissed off at today. ps you can turn this off in the Navigator|Smart Browsing prefs

claudius: Some perspective, please. You are focusing on the cases that Internet Keywords successfully handle, and completely ignoring the *large class* of cases where behavior is decidedly counterintuitive. We all know that this feature can be made to work if you massage it just right. The cases that work aren't the problem. The problem is that when this feature doesn't work as expected, it's *really bad*. Bad enough that, IMO, the good points of this feature do not manage to offset the degree to which it detracts from Mozilla. I'd welcome a correction to the implementation that corrects the UI problems and makes the keyword server configurable. IOW, make the "Search" button mean the same thing as pressing "Enter". But do not expect users to differentiate between searching based on Internet Keywords versus arbitrary terms.

<sigh> I'm not going to argue the merits of Internet Keywords because that not what bugs are for. That is my foremost point. If someone, anyone wants to write a bug outlining the steps that produce some of these eggregious unexpected behaviors and what their expectations were I will be glad to champion their cause. Otherwise, what the heck is the point of this bug?

claudius: I tried twice to divert this to the ng, but people keep commenting here. Oh, well. I have already done what you request: bug 65911, which Matthew referenced in the opening comment of this bug report.

This bug is *not* about removing the ability to do a Web search from the address field. As the summary said, this bug is about the Internet Keywords feature, which (as bug 65911 points out) prevents that Web search feature from working properly. Internet Keywords cannot exist *without* preventing the Web search feature from working properly, which is why it has to go. That's what the heck this bug is for. I refer you to German Bauer's comment in bug 45719: | | The discussion was part Netscape only as *keywords have nothing to do | with Mozilla's build*. | [emphasis added] I have no problem with Netscape including Internet Keywords in their commercial builds, causing a disconcerting user experience for those who thought they'd chosen a different search engine. I don't care. I filed this bug in Bugzilla, not in Bugscape. > When one types two words into the url bar eg 'honda accord' and presses enter > what is it we expect to happen? Pre-internet keywords the answer is 'nothing' > - an error alert. Only since the advent of Internet Keywords has that > behavior led to a search. No it doesn't, and that's precisely the problem. Instead of doing a search for information about Honda Accords, using my search engine of choice, it routes me (via a Netscape server) to a page which no longer exists on Honda's Web site -- giving me no information about Honda Accords whatsoever.

> this bug is about the Internet Keywords > feature, which (as bug 65911 points out) prevents that Web search feature from > working properly That bug complains about the fact that the Netscape servers are queried, instead of the default search engine. As I showed, Internet Keywords are not inherently tied to Netscape servers. In fact, the server that is queried is determined by a hidden pref. You can use Netscape Keywords, Google's I'm feeling lucky, or almost any other web "GET QUERY", including normal searching. So, technically, the only differences between search and "Internet Keywords" are: - They are accessed differently (Enter vs. Search button and Search panel) - One pref has comfortable UI, the other has none at all. - They can (but don't necessarily have to) have different target URLs. These differences allow me to get *directly* to the first search hit on Google (i.e. use the "I'm feeling lucky" feature), when I press Enter and to go to the Google search pages (enabling me to access the second and lower search hits) when I click search. So, technically, Internet keywords save me one pageload and one mouseclick. See my comments on bug 65911 about an alternate implementation of Internet Keywords.

BTW: "Honda Accord" works for me (with Beonex Comm.): I (ultimately) get to <http://www.adnetcomp.com/honda_accord.html>.

BenB: But you're still requiring users to differentiate between pressing "enter" versus the search button, and corollarily differentiate between keyword searches versus "generic" searches. From a UI perspective, both of these should be non-starters.

I see the point now - mpt just wants this bug so that bug 65911 will work properly. I submit that this bug doesn't technically block that bug BUT I do accept that removing the feature would provide for a much cleaner UI experience. Of course, we would be sacrificing what many believe is a useful feature that can be expanded to not be netscape-centric. eg Internet Keywords is just google's "i'm feeling lucky" just much more higly optimized. Just as a note for all who ever consider how much 'enter' should act like 'search' or vice-versa (which means a totally different thing). Just stop and think about what would actually happen and what the user expectation may be in each case: type 'apple', type 'southwest airlines', type 'http://www.mozilla.org' ps BenB: 'honda accord' -->http://www.honda2001.com/models/accord.html w/Internet Keywords (except that honda changed their site)

No, I don't just want this bug so that bug 65911 will work properly. As I stated in the original report, this bug *also* blocks bug 14532, and it is *also* bad UI for reasons other than the silly Enter/Search distinction. > Just as a note for all who ever consider how much 'enter' should act like > 'search' or vice-versa (which means a totally different thing). It doesn't mean a totally different thing, to anyone who's ever used the Enter key on a Web search engine site before. If Mozilla thinks it does mean a totally different thing, then Mozilla is wrong. It's that simple. > Just stop and > think about what would actually happen and what the user expectation may be > in each case: If Netscape chooses not retain Internet Keywords in their commercial tree, then they could quite easily set up `Netscape Internet Keywords' as their default search service for address field searches, and get exactly the same result as they do now. Similarly, Beonex could use `Google's I'm Feeling Lucky' as its default search service.

Taken from my comments in http://bugzilla.mozilla.org/show_bug.cgi?id=65911 As it is I think we provide a decent solution. Some of you agree, some of you disagree. Unless we come up with something that everyone agrees with or you someone can provide concrete evidence that this sux then we should keep what we have. It was based on UI studies and designed by experienced UI designers. No it's not perfect but no one has suggested anything perfect. As it stands we have many ways to disable and configure it so if you hate it then configure it. Here are you options: 1)Turn it off in the prefs. 2)Go to your all.js file and edit the keyword.netscape.com to whatever you please. 3)Build a UI in prefs so you can change the keyword server from the prefs. 4)Build a UI and extend the sherlock parser in search to include a keyword attribute. Then when you change your engine it will also change this pref to the keyword you want. But remember not all search engines have a keyword server. What happens then? 5)For all I care, point the keyword as default to google. I just think the netscape keyword server is a million times better since it does searchs if it can't find the keyword. 6)Turn keywords off by default in the prefs. Keywords are a very powerful thing. Look at AOL. It is build on keywords. We've designed this so you can configure it. I would hate to tear technology out of mozilla because people don't know how to configure the technology.

4 sounds great! 6 can be a temporary fix. 5 what do you mean by "keyword as default to google" do you mean the "I'm Feeling Lucky" feature? I'd like to add few more suggestions: 8) Add pref to use search instead (acts as if search button was pressed) 9) Use IE style %s.com,www.%s.com,%s.org,www.%s.org substitution

or 10) a combination of 8 and 9 ( meaning do 9, if fail do 8 )

From the networking perspective, I think there is a strong case for this to be "off" by default in Mozilla. This would help us expose, understand, and fix a variety of problems we have with name resolution of URLs. It can be on in Netscape, that's a commercial decision.

If multiple search engines were implemented for the URL bar dropdown (which I filed bug 84908 on), Netscape's keyword engine, and Google's "feeling lucky" queries could be generalized to just another entry in the search engines. This is a HELL of a lot more consistant, because right now internet keywords makes very little sense. URL bar: [ www.foo| v] | http://www.foobar.com | | http://www.foofoo.net | | http://www.foo-la-la.fr | |---------------------------------------------------| | [G] Search on Google | | [G] Google's I'm feeling lucky | | [B] Search on Bugzilla | | [N] Search Netscape keywords | | [M] Get map from Mapquest for address | +---------------------------------------------------+

this bug will have to wait around until there is some definitive resolution ofthe issue (which should really happen elswhere - newsgroup anyone?)

Yeah. We need to find some traction on this issue which will probably involve more discussion, not less. Meanwhile, I've filed bug 100412.

Since Mailnews is 20 times slower then Nav on Linux, I'll start the discussion here. How about my UI spec? It makes searching better, AND provides for the direct content location that the keywords do. Plus, it's not tied to any particular vendor (Netscape), and lets users have more then one to choose from. It's also a lot more logical in that it works like the current search does, rather then just blindly pressing enter and expecting the browser to contact /someplace/ and get you /somewhere/.

Question: is it imperative that the Internet keyword feature be removed now? The reason I ask is that there are 2 RFCs about to become standard-track documents on CNRP (Common Name Resolution Protocol). These RFCs define a standard way of doing Internet Keyword search. The hope is that there will be public Internet keyword servers available in the near future and Mozilla can take advantage of them. I believe Keyword search is a part of the future in Search and prematurely removing this item may not be a good idea at this point.

If we are hardcoded to use any proprietary approach in Mozilla, we should remove it. There is probably some way of making this feature more modular. If netscape wants to do something proprietary, it should be in the commercial build only. If there is some future spec coming on this, then Mozilla can support the standard when it becomes available.

*** Bug 65911 has been marked as a duplicate of this bug. ***

This is something that ought to be decided by Mozilla 1.0, I think. I'm not sure how pertinent it is that there may be a standard for Internet Keywords. I'm just not sure it's a good feature. Case in point... Try typing in "beetle" as an Internet Keyword. What do you get? I wound up, not entirely unpredictably, at Volkswagen's Web site. What made that rather predictable? An extensive, decades-long marketing and sales initiative by a big company. The fact that I wound up at Volkswagen's Web site is not simply a consequence of that initiative--*it is part of that initiative*. Now, I didn't ask for "VW beetle", I asked for "beetle". So why did I get a page about a car rather than a page about an insect? After all, the car was named for the insect. *The insect is primary.* Why didn't I wind up at www.coleoptera.org or somesuch? Because keyword listings go to those who can/will pay for them. They exploit naivete (or sometimes just laziness) in users to the advantage of corporate sponsors. Now, is that good or bad with regard to the position of mozilla.org? I honestly don't know if it's either. I just know it's something I personally am not interested in using. Since I can disable it, my needs have been met. Mozilla's proprietors just need to decide who this project serves, and how it should serve them. As to moimoi's suggestion that public keyword servers may become available, I am skeptical. Who would pay for these servers? Who would use them? Who would administer them? Who, if not the almighty dollar, would broker disputes over contested words? In fact, this system is already in place. It's called the Domain Name Registry. We need another mechanism for doing that about as much as I need another hole in my head.

If you don't like it then don't enable it (it will soon be disabled by default in mozilla builds). If you do like it but you don't like netscape's server results the customize it. I point keywords to google's I'm feeling lucky or a plain search engine.

Pretty much anyone who has a web site can make it a keyword search. We are documenting that as we turn off the keyword defaults in the mozilla builds. I don't think we will ever remove this feature, the feature has an on-off switch + it can use any server that is properly set up.

I don't see any advantages to a "keyword server" over a search engine that redirects you to the best match, like Googles "I'm Feeling Lucky". The only differance is, using the search engine hookup we have in place, it's more uniform and less bloated. If you could select multiple search engines, and a default one to use when you've typed in, say, "foo bar" and just hit enter, it would be more flexible, and a lot less confusing. Netscape could use this too. Netscape search and Netscape keywords would be the two search engines, with NS kw as the default. just redo your backend server to work like a search engine (it probably does already, I dunno, never used it).

> The fact that I wound up at Volkswagen's Web site is not simply a > consequence of that initiative--*it is part of that initiative*. This is a valid and smart argumentation, but not for this bug but for replacing Netscape's keyword server with another "keyword" provider. You're basically arguing like saying that Microsoft and Netscape (ab)use browsers to push their interests, so all browsers in general are bad and should not exist. > If you could select multiple search engines, and a default one to use when > you've typed in, say, "foo bar" and just hit enter, it would be more flexible, > and a lot less confusing. That's what you can do today with Google. I won't be able to do it when this bug is "fixed".

No, Ben. I didn't say anything about Microsoft or Netscape. What I'm saying is that Internet Keywords, as a feature, is designed to favor the interests of advertisers rather than the interests of users. Google's "I'm feeling lucky" is little more than a novelty. And who's to say that they won't one day let advertisers pay to be "luckier" than others? You suggest that Microsoft or Netscape are unsuitable as defaults, but by what rationale would you suggest that mozilla.org select a default?

> That's what you can do today with Google. I won't be able to do it when this bug > is "fixed". That all depends on how it's fixed. Making keyword servers just another search engine is a lot cleaner, and gains functionality, it doesn't lose any. I haven't see anyone put forth any reasons why "keyword servers" or this "Common Name Resolution Protocol" are nessicary to have as a seperate entity. All it does is limit flexibility, and make the prefs confusing.

> What I'm saying is that > Internet Keywords, as a feature, is designed to favor the interests of > advertisers rather than the interests of users. No. Technically, Internet Keywords is a kind of search, that directly forwards you to exactly one result, instead of showing you a page with results. It is no more or less right or wrong than the search (Sherlock) feature. > And who's to say that they won't one day let > advertisers pay to be "luckier" than others? Unlikely, because their success comes from them not doing so. Even if so, there are more search engines out there, you'll probably find one that pleases you. > by what rationale would you suggest that mozilla.org select a default? Because every pref has to have a default? I.e. by the same rationale by which the search feature has a default.

> Making keyword servers just another search engine is a lot cleaner That's would be paprt of bug 65911, as I understand it. This bug is about completely removing the feature, without replacement. > I haven't see anyone put forth any reasons why "keyword servers" or this > "Common Name Resolution Protocol" are nessicary to have as a seperate entity. The disadvantage of a normal search is performance/resources. If you do a normal search, the search engine has to create 10 or so results, and the browser has to parse a normal HTML page. (HTML is not the ideal medium for structured data.) I am assuming that you want to keep the way the feature works today.

>> I haven't see anyone put forth any reasons why "keyword servers" or this >> "Common Name Resolution Protocol" are nessicary to have as a seperate >> entity. > The disadvantage of a normal search is performance/resources. If you do a > normal search, the search engine has to create 10 or so results, and the > browser has to parse a normal HTML page. No, you didn't understand. I'm only proposing the same interface is used. You wouldn't get back 10 results from the search engine, just a 302 temporarily moved. For example, if this were implemented, I would select bugzilla, google, google's I'm feeling luck, and freshmeat (software site) as my search engines (that's four). I set google to default. If I type in stuff that's not a domain and hit enter, the default is used. The other four are available in the drop down. For Netscape's browser, they would have both NS search, and NS keywords selected, with NS keywords as the default. Any "keyword" server could be integrated this way. It's just a differant query to a search engine, really, so let's treat it as such.

> It is no > more or less right or wrong than the search (Sherlock) feature. I never said it was right or wrong, Ben. I'm simply pointing out that by working the way it does, keyword search surreptitiously puts a lot of control over content promotion in the hands of the keyword server. > > by what rationale would you suggest that mozilla.org select a default? > > Because every pref has to have a default? I.e. by the same rationale by which > the search feature has a default. You misunderstand. I'm asking you what you think the default should be, and how you arrive at that conclusion.

> I'm asking you what you think the default should be, and how > you arrive at that conclusion. That's offtopic. File a new bug and I'll answer publically there.

I'm not sure what the summary of this bug should be. Personally, I'm in favor of having both internet keywords and search. I have keywords set to google and search set to bugzilla. Very convenient. It sounds like the consensus is to keep internet keywords, but maybe change what happens when it's disabled (so that a search is triggered when IK would have been triggered had it not been disabled). Is that so? Can we change the summary to something like use 'search' instead of internet keywords when IK disabled Also, for those interested in internet keywords going to a real search engine, not that currently you can't send arbitrary queries to your search engine via internet keywords. Any query with a dot character will abort an attempt to use internet keywords, even when it's enabled. This is bug 79655 and has a two-line fix. Anyone want to walk me through fixing it (as I'm a bit low in mozilla patching clueidity), or do the tiny patch yourself? (patch described in that bug). -matt

> use 'search' instead of internet keywords when IK disabled I think, the idea of the Internet Keywords pref was to not interfere with normal URLs. I.e. make Enter always load the URL as typed.

That would be a separate bug, because it would be a mode (hopefully available via a pref). I would like to resolve this bug as WONTFIX, because it seems clear to me that nobody has come up with a good case for removal of the IK mechanism. There are aspects of how the IK mechanism and search are similar, and both need improvement, and there are conceptual or implementation aspects that are confusing a lot of people that hopefully will be cleared up over time. For example, I am seeing a lot of comments about how a Internet Keyword system is bad because it is a controlled namespace (I call this "the tyrany of the namespace". This is a problem with the way people construct, use and share public namespaces, but is not a reason to remove the feature. Also, I am seeing a lot of comments about how search and IK are similar different. At the basic level, they both take a string and forward it to an HTTP server, so in some ways they are the same. I think that on an application level, there is a desire to have IK servers to reflect some form of 1:1 relationship, while search is a "list of" result. (In a perfect perfect world, search would always return what you want, right, which would be the single perfect destination for you and you would only get a list back if you said "show me all"). On a user interface level, mtp has made some good points about URL bar UI, which I don't understand as well as the networking aspects. I will say that in my mind, IK should be a tighly coupled service (one entry point, one server), where search is a potentially parallel conecpt (so the UI should allow configuration and easy access to multiple repositories).

> nobody has come up with a good case for removal I stand by commnet 36. > I think that on an application level, there is a desire to have IK servers to > reflect some form of 1:1 relationship, while search is a "list of" result. Why? Why should a user be allowed to have multiple search engines that return "lists", and only one that does the redirect deal? They're the exact same thing, having two mechanisms only greatly confuses the matter. And why should the default (what we look up non-domains with) only be a redirect type, and not a list of? Why shouldn't the user select. The above does assume support for multiple search engines will be added, but I assume that's a no-brainer. As-is, we're no better then IE's built-in search functionality, but with the use of the URL bar dropdown, we have much more potential.

If you think someone would actually use multiple intenet keyword servers, you can file an RFE for that. My thinking is that the IK's should represent a single user-namespace (the server could always amalgamate other namespaces for the user anyhow). If you do think this should be done, you still need internet keywords, hence this bug is invalid or wontfix. As for multiple search engines, this makes sense if you don't believe that the one-size fits all search engine trend will last forever. I'd like to see more focused search engines (an English diectionary, a PERL syntax guide, etc).

Folks listening here might be interested in a bug I filed, bug 128857 - "Internet Keywords search should use Default Search Engine". I contend that the UI for selecting an IK server is already here, it's the "Default Search Engine" selector. If the user wants to keep the current behavior, just set Netscape Search to be the default and the results remain the same (for all the tests I tried). I have no problem with that being the default - hey Netscape pays for alot of work on Mozilla, there's nothing wrong with throwing them a bone, but if you have lots of Anti-Netscape mojo going on, it's a UI-accessible change away. This method also reconciles the 'two searches' confusion problem. It might be nice to also have a dependable suffix added to the query string, e.g. '&mozilla-ik=true' so, as someone writing a search engine, you can know you're handling an IK search and do a 302, if appropriate. I'd add this to the URL field if I had perms. http://www.mozilla.org/docs/end-user/internet-keywords.html

See, that wasn't a very original suggestion. Internet Search already does something just like this, it adds: &sourceid=mozilla-search to the end. If we just have internet-keywords do: &sourceid=mozilla-keywords It would be very symmetrical, and probably alot of code reuse could be had.

bill: #43.

But #43 suggests multiple IK servers, I'm only suggesting using an alternate IK server, and a way for Netscape to continue to provide IK service with a unified IK/search approach. When you say: >My thinking is that the IK's should represent a single user-namespace Do you mean "Mozilla should present a user with a single IK namespace, but the user can pick where he gets that IK namespace from" (I agree) or "Mozilla should present the user with Netscape's and only Netscape's IK namespace" (I disagree). My approach allows anyone to provide an IK namespace without Mozilla having to maintain a parallel IK plugins collection. Again, I appreciate Netscape's support of Mozilla and I don't think having Netscape as the default search/IK is a problem at all. I'd hate to see IK go, but if this feature has to force Netscape content on people (very unusual for Mozilla), I'd have to reluctantly agree with the reporter's original sentiments.

These bugs are too long, and too focused on legacy, psuedo-marketing terms. Somewhere in one of these bugs I have said: "The only difference between IK and search is that IK is on by default, and points to a single server. Search is not on by default, and is a configurable list." The rest is irrelevant, because the logic is only server side. You aren't actually proposing anything that probably isn't already available (but undocumented). This bug is actually for the "absolutely tear it out forever" people. The make it a pref bug is somewhere else, and enhancements are discussed separately. There is also a bug for turning everything off by default. I think your comments actually belong in one of thos bugs. I guess now is the time to say post 1.0, I'll probably codify some terminology, write documentation, and freeze all existing bugs, and restart new bugs in bugzilla, because there is too much drift.

>"The only difference between IK and search is that IK is on by default, and >points to a single server. Search is not on by default, and is a configurable >list." That would be great if it was so, but one of my interests in IK is space-separated search terms with the enter key in the URL bar - IK is the only way to that functionality, per bug 58867. >This bug is actually for the "absolutely tear it out forever" people. The make >it a pref bug is somewhere else, and enhancements are discussed separately. >There is also a bug for turning everything off by default. I think your comments >actually belong in one of thos bugs. OK, I'll try to find the right one. Still, in the reporter's original complaint he concluded that IK "appears to exist solely to drive users to Web sites of Netscape and its partners" and I was trying to address that concern.

OK, since this bug and bug 53171 are just sitting here, I'll propose a design to fix them. We currently have way too much stuff that does relatively the same thing... Netscape keywords, search engine prefs (with UI), and custom keywords. Ditch em all. All will be done with sherlock/whatever-we-call-them-now plugins. This fixes: Silly limitation of 1 search engine Silly limitation of 1 keyword server No UI for keyword server No easy UI for making custom keywords No way for web pages to install keywords--with user approval, like search eng Very confusing way which enter works when combinations of search/keyword prefs are set No useless custom keyword bookmarks that would load a URL like www.site.com/%s if used directly Internet Search pref pane would do something like: +--------------------------------------------+ | [X] [M] Mozilla Bugzilla ^| [ Add ] | [X] [G] Google Web || [ Remove ] | [ ] [G] Google News || [ Edit ] | [X] [G] Google I'm feeling lucky [default]|| [ Set as Default ] | [X] [N] Netscape || | [ ] [N] Netscape Stock keyword v| +--------------------------------------------+ The [X] means the checkbox is selecting and the engine is available. The [M] etc is just the engines icon. Selected/active engines show in the URL bar drop down (if that pref is set). Non selected engines still have their keyword available. Add would pop a dialog allowing you to insert the URL of a sherlock file, or enter in information yourself (much like currently creating a custom bookmark) Remove removes an installed sherlock plugin from the users profile. Edit would let you change things about the selected plugin, like it's keyword. This interacts with the URL bar as follows: foo [enter] attempts to DNS foo, and if it doesn't exist, loads the new error pages that are being worked on (which lets you search for 'foo' with a click). ? foo [enter] searches the default engine for 'foo' foo bar [enter] anything with a space and no keyword searches the default eng. keyword foo [enter] uses a specified sherlock plugin to search for foo Search engine plugins are stored in the profile directory, not in the user profile directory so they don't need replacing every time the browser is upgraded. Optional second step: Google and Netscape are evangelized to offer the 1-click install for the plugin to Mozilla/NSCP users on their pages, so we can remove all 3rd party search engines from the base distro. Any the user actually wants they can click to install. This would unfortunately mean search by default ends up searching data a normal user doesn't care about (bugs, bonsai...)

Jeremy, just to reiterate, this bug is for the "absolutely tear IK out forever" crowd. You should file a new RFE bug for your suggestion.

Bill: IK, really a generic keyword system, since the URL is editable, will not be "torn out forever" until its (valid) functionality is replaced by a superior system.

No arguments here - I wrote a proposal to enhance IK in bug 137026 - but there are some whackos [as a term of endearment] who think IK is evil. See comment #1. My only point is that they're welcome to have these ideas and this is their bug. Let's get your proposal its own bug and discuss its merits there.

I use IK to search without hitting the search button. I have configured Mozilla to run a Google search (not I'm feeling lucky, but a normal search). Now I can enter whatever I want into the address bar and if the DNS server returns no result for it, I get the google search for it. That's a great feature.

(I sure hope someone reads this) This "feature" should be removed. The whole INTERNET KEYWORDS INFRASTRUCTURE IS FUNDAMENTALLY FLAWED BECAUSE IT IS VULNERABLE TO SPAMMING OF SEARCH ENGINE RESULTS. If this feature is left in and is defaulted to "ON", I feel it will be a huge loss for open source. And keyword spammers and pornographers will almost certainly use it to their advantage. For an example of what I am talking about, just type in "localhost" in your browser (make sure your local web server is not running). The default Google "I feel lucky" search redirects users to a multi-level pyramid marketing scheme. There is, unfortunately, no way around this - if Google cannot figure out how to stop keyword spamming with their army of smart engineers, I don't think Mozilla project should try to tackle this issue until the engines can be relied upon. Please PLEASE keep in mind that the only reason this "feature" was included by Netscape was so that they could control (and charge advertisers for) keyword traffic (since it was going through their servers). Now that Mozilla browser is open source it should be taken out. Users should be free to make their own decisions about where they want to go. Don't you think if a user wanted to go to Google and search for a topic, that they would have done that? It is presumptuous in the extreme to decide where to send the user. Because it was originally designed to shift traffic to Netscape, THE IMPLICATIONS OF PROGRAMMATICALLY REDIRECTING LARGE NUMBERS OF INTERNET USERS HAS NOT BEEN THOUGHT THROUGH. "localhost" is the most obvious example of this. Now that Mozilla browser has passed IE in terms of usability, huge numbers of people could be adversely affected if this "feature" is not removed. I sure hope someone reads this...:(

Resolving as invalid, since this bug report is now too long and content-free to be useful to any programmer. I am still of the opinion that the distinction between "Internet keywords" and Internet search is arrant nonsense (Bugzilla and Amazon search plug-ins, to pick just two examples, could take you directly to a single result for some searches just like Internet Keywords does). But I suggest future efforts to remove the misfeature should involve (1) making a patch to remove it then (2) filing a bug report, not the other way around.

RESOLVED: won't fix. This is off by default in mozilla, but there is a lot of interest in making this work for FireFox, so I can't see this being removed anytime soon. Also, we have documentation and this is pref'd, so I think users have the proper choices (turn it off, or change your IK URL if you want). There are also some interesting product futures for making this work better, as well as fixes (frankie: IK does not call to localhost spuriously in newer builds).

> there is a lot of interest in making this work for FireFox Plus, I need it in my products. Comment 55 is just off, this is a very useful feature, esp. when compared with the alternatives like domain guessing. WONTFIXing based on mpt's and benc's comments.

Just curious, with the departure from Netscape/AOL, who's maintaining and funding the keywords?

(In reply to comment #58) > > there is a lot of interest in making this work for FireFox > > Plus, I need it in my products. > > Comment 55 is just off, this is a very useful feature, esp. when compared with > the alternatives like domain guessing. > > WONTFIXing based on mpt's and benc's comments. Well, I'm glad it got read at least...:) Just out of curiosity, who is this "feature" useful for? I also don't see how it can be "necessary" - it is definitely an ancillary feature, at best. While I can see how it would be intellectually stimulating to write code that correctly anticipates user desires, I still don't see how it provides benefit to end users from a practical standpoint. It takes microseconds as a user to go to Google and type in a search - how does having the browser preprocess every URL string to try and determine if what I want to do is search help me as a user? Especially if the code is not guessing right 100% of the time? What about those situations where I would like to know that the location I am trying to reach is unreachable? What if the server I am trying to reach is down? Will I get redirected? What if I just pushed a new site live and DNS has not propagated fully yet? Will I get redirected somewhere else? Also, in a real world scenario, a system that decides where a given string will send the user could be misleading and confusing if there is no indication or feedback given that the system has done a search or redirection... My first thought when I saw the redirection was that my PC had been infected with a virus - I was surprised to see that the internet keywords concept was still around (I thought it had died with RealNames). It took some looking around before I was able to see what was happening. I would bet that most users are unaware that this redirection is even going on. Also, any system that relies on a third-party search engine like Google is going to be vulnerable to spamming... how are you intending to deal with that? What if my competitor hires a company to spam Google and get top placement for my company name? Is every Mozilla user who types in my company name going to get redirected to my competitor? How do you feel about millions of businesses that are currently not spamming search engines STARTING to spam them because Mozilla's inclusion of this feature makes it a necessity in order to remain competitive? As Mozilla's share of the browser market increases, someone is going to have to think seriously about these issues...

To my understanding, Internet Keywords only intercept input which is *not* syntactically valid as URL, e.g. stuff with a space in the middle, see comment 40. This is useful to enter "Mozilla Firefox" or "? Firefox" and get directly to the right page, for most words (it does work for "? Firefox", just tried it in Beonex Communicator - not sure how other products are configured). So, it's useful in all cases where you'd want domain guessing (like that Firefox feature), but much more correct, and in many other cases as well. Loading Google, entering the search term, clicking Search, clicking the first search result takes *far* longer. Even if you use the built-in search feature, IKs are still faster.

Verifying based on comments and request of the bug reporter, based on 1.7a behavior. A few notes are in order, based on recent unfounded complaints: 1) This is off by default. Only people who want this behavior get it. 2) Mozilla gives a list of seach results from Google, not an I'm Feeling Lucky 2a) this isn't a Firefox bug 3) The keyword server is configurable; there's no UI to do it - somebody might want to file a separate bug on this. 4) IK only operates on invalid URL's - this will never touch a valid request URL - even a bare "localhost" will try to contact http://localhost:80/ . 5) IK does not fire on DNS failures. I tried several invalid hostnames, with and without my proxy server, and I never got IK invoked, just an error dialog. People are making up problems that don't exist here....