The Discussions on Traffic Reform with the Lindens

Can't you just use the trafficbots?

They can serve double duty

It's much more than the old vote box. The vote box was simply a count of votes, but the links-based All search not only evaluates links, but it also evaluates content, and the way that everything is evaluated is complex.

[added]
Pickbots aren't really feasible. You can't just add a Pick to loads of bots and expect them to count - they won't.

[added]
Have a chat with Tiana. She followed the information in this forum and reached #3 on page #1 for her main searchterm, and she now gets plenty of money from her store

The phone book is a list of names, those looking under Z are looking for a name starting with Z.

You don't have the yellow pages where you live?

The problem with alphanumeric listings is that they are only useful for finding specific places - like the phone book. Nobody is going to click through the pages to find a place that's name starts with Z, so it wouldn't be anywhere near as useful as the phone book, unless the start initial could be specified. BUT, if you know the name of the place you're looking for, it can be found easily in the All search, so there's no need for a seperate alphanumeric directory listing.

The idea of random starting initials is thinking of an alphanumeric directory as a search system, which it could be used as, but the places that get listed in easy sight are random, and may be no good for the searcher at all. Another consideration is that people using search on the web expect to be able to find what they found yesterday by using the same search, but if the start is random, they couldn't do that.

We are talking the places listing here - Which is supposedly older and obsoleted by the new relevancy based search.

Thus why is it necessary it has any other purpose than a list of places?

Yes if such a system were in place you should be able to sort by letter and by categories.

The complexity of the formula is a bit beside the point. People simply put their friends in their picks. They really shouldn't be factored at all. They already serve their purpose in people's profiles. If I search for skins I probably want to see places with the widest selection, not the skin maker with the most friends or the most coercive strategy for getting people to add them to their picks. That has nothing to do with relevancy.

Whether Picks should be used as IBLs or not is a matter of opinion. Links-based search engine treat IBLs as unbiased votes for a webpage, and they manage to produce excellent results by doing it that way. I can't think of a better unbiased vote in SL than someone adding a place to their Picks.

It's been a few months since I brought that problem up, good memory . The difference between then and now is that you ask for a reason to keep places around and that is one that has nothing to do with ranking.

It's more than a few clicks, Places is instant result, on a good day All takes 30 seconds, on a bad day over 2 minutes or I just give up on it.

And again, the in-world browser is broken for a lot of people. Even if you assume that it's more relevant, an "inferior" search that lets me browse all results is better than a "less inferior" search that only shows me 4 results because I can't scroll down to see the rest.

Search for ETD in Places then search for ETD in All > Places. ETD ranks third in the All > Places search which requires you to read over the results and pick the one that you really asked for.

Places has a definite use for me and it's irrelevant of traffic since I'm only interested in finding parcels I already know. All > Places is a lousy substitute for that use of search.

Sure, because web links tend to link from one site to another with related information. If a large number of sites link to a particular site it's a good metric for the quality of information contained there. It's a completely different dynamic at work than with profile picks. It's not even apples and oranges. It's apples and lampshades. Profile picks operate under exactly the same dynamic as the old vote boxes did, and those were done away with because the data they produced was worthless. Many people don't even use their profile picks for places they like. They put favorite snapshots there of their friends. How is that a useful search metric?

Edit - to put it another way, if I go to Google and search for "espresso machines" I want to see the pages with the best selection of and best quality information about espresso machines, not the espresso machine maker with the most friends on Facebook.

Look at the results for All > Places > ETD again... why does Nyte'N'Day rank higher than ETD while it's not even a keyword on that listing?

Ages ago (I'd have to guess 12-18 months ago) ETD was located on Couture Isle, which is Nyte'N'Day's sim and apparantly enough picks exists still that link ETD to Couture Isle, effectively "google bombing" ETD as a keyword.

In this case it's entirely accidental, but can very easily be done deliberately as well which is a new form of search manipulation that's insidious because it's not going to be obvious to people why it's happening.

OTOH, intitle:etd returns exactly the right results, suggesting that the problem might not be so much the search engine as the interface provided in the SL client.

Well, my idea would involve more like a group, with varying different views, who can't review their own properties, and may even be limited to people willing to do it, without *having* any properties to be reviewed (which would be optimal, but not likely). And yes, part of the major problem we have at the moment is the Porn industry model of key words spam salad. Throw a bunch of stuff at a wall, often without any correlation (like neko and tail, which may mean they sell neko eyes for people looking to get tail...), then hope some poor sap is looking through the holes in the wall at the right time to get hit with the resulting feces. If you don't limit descriptions to **real** descriptions, or create a keyword system that isn't derived from just throwing a salad bar of meaningless words at a blank wall, you get the same kind of result you expect from going to some place like Yahoo, selecting to allow it to show adult content, then entering something like "britney spears pants", which, even in google, gives you, since I just tried it, news stories and photos of her dropping her pants in public, while with Paris. I am *real* surprised it didn't come up with some laundry list of, "BSDM britney pants sex money speared blah blah blah blah...", as the first thing, but then Google at least tries to be less stupid about that. It ***would have*** if it was SL and using fracking traffic data, and no way to refine the searches to eliminate places which had pictures of her, instead of designer genes based on her clothes, or some such, which one would **presume** would have been the intended result for most people, even *with* "show mature content" selected.

So long as the search is based on the word soup added to the description, and not on a more specific, clear, and non-gamable metric, the dishonest people will just spam the description field, then count on traffic to push themselves up to the top ten, or even 100.

Imho, we badly need a "basic", and free, means to designate the primary use the property will have, preferably with some way to specify who else is affiliated with that property, like when dealing with malls, and leave the pay ads to cases where you want to be more specific. Stop letting people just throw garbage at the user of the system, without any way to filter it out, or clarify what they wanted to find. That can be done "by" Linden, with a drastic rework of the system, or by someone else, who won't provide "any" of it for free, using something like what I suggested a bit ago. (Message #224, which no one commented on at all, instead deciding to rehash the same issues with the existing system, which we *all* know is broken and doesn't look to be unbroken by 'All')

I used it as a non-hypothetical example though.

Change it to 'skins', 'hair', 'animation' or any regular other keyword and you can propel yourself to the top with fewer picks/incoming links than you'd normally need for that specific keyword.

intitle: as a default prefix would work, but then you'd loose the description field and "Items for sale".

I've been watching a movie and now I'm off to bed, so I'll save the responses until tomorrow - when you lot are in bed But I wanted to post a quick reply to this part of Chip's post.

You may *want* to see pages with the best selection of espresso machines, but if Google only provides you those, it's done a *very* poor job. From that searchterm, it has no idea if you are looking for pages like that, or pages about how espresso machines work, or manufacturers of them, or whatever. You will undoubtedly find some pages that you want on the first page of results, but you'd need to learn to use better searches if those pages are all you want.

Also, you may not want the pages with the most links from sites like facebook, but that's largely what you'll get in Google.

Just wanted to write that before I go to bed

Ok. but an intitle: search is for when you want to find something with that phrase in the name. Of course it would be the wrong tool you wanted to find things related to that term rather than named after it.

And that's what i mean when referring to interface improvements. The geeky little Google syntax that makes really accurate searches possible needs to be exposed and demystified.

This is a problem with "real" Google too, it also can return crappy results for naive keyword searches, though it does have a certain advantage from the sheer volume of its raw data. And that's what could be exposed better in the interface, that there simply is no one best way to do the search. The new search doesn't have menus to help people refine searches if they don't know the magic google-fu language. There isn't even a help link to point at the magic language!

I saw in on of these threads some remarks that everyone knows how to use the phone book, but schools used to teach that skill and others like it (Do they still? Don't know). Many, maybe most, of the differences people are seeing in results quality comes down to the same issue, tools are less useful if we don't really know how to use them. That problem actually applies to the old search too, each of the places, classified and events tabs have different matching rules that are only exposed through trial and error.

The yellow pages returns a lot more that twenty results per page, it also has one hell of a lot of categories.

No it isn't a bit beside the point. It's true that some Picks are changed and used for personal things, but in changing the words in the Picks, it stops them from being a match for a searchterm, so they don't add to a place's relevance, or improve its rankings, for its searchterms. That problem takes care of itself. There is no better voting in SL than the voting of unchanged Picks - they are perfect for a links-based search system. When handling a query, the system looks for matches in the links texts (those Picks). If people have changed the wording in the Picks for a place, they simply don't match the searchterm, and aren't counted. Perfection

And someone else might want to see a wide range of places that sell skins. One individual's desire isn't everyone's desire. Coercing people to add a place to their Picks doesn't devalue or spoil the results in any way. For example, I sell low prim furniture, and I have a sign in my store that says something like, "If you like this store, and would like to help others find it, please add it to your Profile Picks. It helps in the search results". I've heard that other people pay to have their places in Picks. The point is that the places are relevant to their names and descriptions. They belong in the search results for the things they sell, and it's nobody else's concern how the Picks came about. It's only other people's concern if the Picks target wrong things - like me targeting skins when I don't sell them. If you do a search for low prim furniture and end up in my store, you won't be disappointed, even though I encourage people to add the store to their Picks.

See my previous post.

Yes, the problem of old Picks has been discussed before, and does need to be dealt with.

And the group would consist of volunteer residents - people who like the get involved. Many of them would have huge biases, just like we see in the forum - forum users are people who like to get involved.

Well, yeah. You would need a) someone in control that would kick off people with *obvious* bias, and maybe even a rule that b) the same person can't "review" the same area twice. Its not necessarily the solution I would pick, given an option. I like my idea of limiting the number of "terms" people use, so you don't have word salad, and thus getting more "specific" categories. I look in the yellow pages for Television, I don't *expect* to find taxi cabs in there. If I am looking for a taxi cab *with* televisions in them (as a silly example), I then have the choice of calling all the taxis and asking, which is what you get with the, "teleport to 50 places to find something", issues, *or* you allow some "limited" ability to specify what they have. Right now... You can't even type in a term like RP and **get** what you intended. You get stuff that says, "no-RP", other stuff that just sells to it, still others that cater strictly to sex RP, etc. You couldn't even add "furry" to the list, since a lot of places "allow" furry, but don't list it, some exclude it, and still others support it, but don't explicitly say so.

Either the search system has to categorize things *clearly* for people looking, instead of what you get now, or... something else has to replace it. One of the few things that could is some sort of review system. And that is *only* a problem if you are a) lazy about controlling your reviewers, b) don't care, or c) don't have checks, balances and safeguards in place to discourage bad behavior. Simplest one is, if you can't do your job right, you can't be in the group that does the reviews. Duh!

Any ranking system that depends on the activity, profile picks, groups or votes of SL avatars is a nonsense.
If each unique avatar were a unique RL person, then those indicators would be useful.

If I wanted 10,000 avatars to put me in their Picks, I could either create accounts or rent a pick off someone who had the accounts.

I detect a lot of confusion in this thread (and others, and in-world) about how Profile Picks are intended to work as IBLs for the new Search engine. I think this is because we're all so hypersensitized to the hideous mess that Traffic has become as a Search factor that we're inclined to see everything as "gaming the system." But you can't "game" a system that's explicitly designed to be manipulated--and that's the whole *point* of using Picks as IBLs. The crucial thing to understand is:

Profile Picks are not popularity ratings.

It's emphatically *not* like traffic. It's simply a means for those with listings they want to show in Search to control what text will be indexed for finding those listings. That's all.

Some businesses are engaging in weird, superstitious, magical-thinking *attempts* to game the system by providing incentives for lots and lots of residents to link their Profiles to the parcel listings, apparently on the premise that if some are good, more must be better. It doesn't work that way. Indeed, the more Picks out there linking to your parcel that you don't personally control, the less you're able to control what text is actually used in the indexing.

There's lots more to be said about this, but a tremendous amount has already been explained in the threads quoted in the forum sticky on this subject. But there's still too much misunderstanding about how Picks work, and their intent as IBLs. It's not that this is the best possible system for generating IBLs, nor is it very analogous to how IBLs get formed when Google search indexes the web. But it's just not sensible to rail on against something for not being "fair" at doing something it was never designed to do in the first place, or being "gamed" for being used exactly as it was intended.

If you type 'RP' into the search box, the results *must* contain the things you stated, or it would be a very bad search system. *You* may want only of them, but others will want the other things. What you personally "intended" doesn't matter if you search on such broad searchterms. No search system can read minds.

Search systems don;t categorise anything. They just return relevant results for the searchterm. A directory syste, as you propose, categorises things. It's not a bad idea to have a directory system in SL, but it would be no good at all as an alternative to an organic search system. It requires editors (reviewers), and they would be unreliable in various ways. It would soon suffer huge outcries from people due to the huge delays in reviewing places, due to the certain miscategorising of many places, and due to the thing being out of date in no time at all - places moving and closing but the changes not reflected in the directory for a long time, causing people to TP to places that are no longer there or have changed.

Apart from small niche directories, directories have never worked on the web - Yahoo! being the one exception. There are huge problems with non-paper directories, and if LL ever created one that actually got used, it would be a huge failure, just like all the web directories.