The Discussions on Traffic Reform with the Lindens

Perhaps so, but does that mean that the place shouldn't rank highly for the searchterms, or higher than those it overtakes? Of course not. What if it's the best place in SL for what it does?

There's an RL parallel to it. Since Google came along and introduced links-based engines to the web, people have been arranging links so that their pages rank higher than they otherwise would (Google made a pigs ear of the natural web). Not long ago, they decided to attack some of it, and they placed paid-for links as being against their guidelines and subject to penalties if found. LL could take a leaf out of Google's book on that score. It'd only need one announcement in the blog.

And should, but it would be easier just to stop factoring picks or traffic at all.

Not really. All it would have left is page content, and that's much easier to 'organise' than Picks. Don't forget that the vast majority of Picks are genuine, and should be counted as genuine votes for the places. And such votes are a far more reliable source of ranking data than page content.

Yeah. It's just that I'm guessing somebody has affect tied up in the lame idea of using Profile Picks text to feed the engine, and now there's developing a whole constituency of residents who are investing in making that work for them. So, it will be some effort to pry this loose, or to redefine how it's used.

Just for idle amusement, the pre-GSA jira I mentioned is http://jira.secondlife.com/browse/SVC-633, the idea of which may look familiar to anyone doing Information Retrieval back before Google "solved" that problem. (I swear: any day now somebody is gonna decide Codd was all wet and we should just replace First Normal Form with freetext Google queries. )

This would be what I call 'another' bad idea. Already, there are things in place to form a sort of barrier to people who joined SL long ago. People who own land, but are not allowed to own land under the current rules, and people who have legitimate problems they wish to have addressed, but can't because they're not paying for the privilege.

Restricting the counts to only people who have premium accounts would reduce counts in some areas to zero... even though those places would be PACK with people. How would this actually reflect anything other than a rather huge unconcern for the people who play SL, but don't PAY for it?

I can't say that I understood all of that Jira, but I did recognise the word "entropy", and, just like web engines, whatever system is used in SL will comply with the law of increasing entropy

We've been discussing Picks and their effect in some detail since the post that you quoted, Qie, and I really don't see Picks becoming the bad thing that traffic became. If there's a chance of it, all it would take is a blog from LL that paying for Picks is outlawed, and that would deal with it.

Is there really no way that LL can detect if someone is online or not as a bot or a normal playing character? Like if they were to be online for three or more hours a day, never moving, using any game features, IMing or chatting or anything? Isn't that something that can be detected and thereby used as a way to detect bots?

To me, then problem isn't about picks or about the actual numbers you see for a sim or location when you search; it's about the load (or absence thereof) on the servers and on the bandwidth that SL has or uses. Do bots cause trouble in the accessing and use of the servers? I would think the answer would be at least 'somewhat', as when you go to a sim with these bots in them, they are wearing objects and standing around, making you have to deal with rezzing them and their textures. The few dens of them I've found have basically never fully appeared, due to all the textures and items they had on them.

As for the consumer, 'trying to find a place' aspect of it, I did a recent search and ended up going to two places looking for something that wasn't in either place due to boosted numbers. I had to eventually skip down the list even further before I found a place that had what I was looking for.

This isn't "the web", we are not dealing with web pages, where you are hunting for "color of spiro agnews underwear", and don't mind getting quotes by him, or just colored underwear. This is a simulation of a real world, where the only thing we *don't* have in it is street addresses (thought location SURLs are similar) and fracking phone numbers.

A search system for web content that tries of categorize won't work because a) your dealing with blind systems to do the categorizing, b) there may be an unlimited number of things *on* the pages, and c) its too big a pain to designate what category you "should be in", because, well... The only may you can do that is with something like "meta" tags, and those can be gamed, which is the problem we are already talking about.

What we have here in SL is different. We are dealing with a system that, basically, has a) limited numbers of things that "can" be on a parcel, b) often clear content for those places, and c) the ones that **don't** are the equivalent of flee markets, shopping malls, or Wal-Mart. Someone wants to "look for" a mall or Wal-Mart in RL, to they throw up their hands in despair, or look in the phone book for "mall", or "Wal-Mart"? If they instead want to find Radio Shack they *look for* Radio Shack, and that gives them a list of addresses, one of this may **gasp!!!** be in the mall, where they can also shop for other things they needed.

Mind you, it gets a bit fuzzier when looking for something like "batteries", since one needs to already be aware that those are "sold at" both Radio Shack and Wal-Mart. But, that is where a pure search gets useful, when available, and you *expect* when you enter something like "furry avatars" to *get* places that sell those, without having to slog through pages of stuff that happen to contain both terms, but not directly associated with each other.

Point is, yes, a directory wouldn't probably work for "everything", its only useful to have it for the "top level" searches, the minutiae of just what "precisely" you are looking for would have to be left for a subsearch, on the actual products.

So... Lets say you have widgets, and you want to find a wizbang boogle widget. You do a main search:

Gadget's Widget Shop
Fraznagles Goofy Gadgets
Fred's Mall

You then do a "search within these locations", or some such, and find that both Gadget's Widget Shop, and Fred's Mall have them. Now, if you then go back and look at Fred's Mall, you find they also have other stores you want to look at so that is where you go. Otherwise, you might go to Gadget's directly.

On the other hand, if you just want to find "Gadget's Widget Shop", you may get a list like:

Gadget's Widget Shop
Fred's Mall

To me this is just Duh! But, it requires that you limit how much word salad, if any, someone gets to throw at the search system, and that there is clear information of just who is being hosted "in" the malls. Self-reporting won't work, since obviously the guy running the mall can forget, or just not bother, to add Gadget's Widget Shop to his "list" of kiosks. Therefor, imho, you need something that the "seller" can set to do that, since its easier for them to go back and fix it anyway. And it could even be automatic, if the right script was in the vender. They just have to remember to set the vender to report itself as theirs, which is has to do anyway, to the purchase system, so, it registers itself one more way, so that the region's "category" system knows what is going on.

Yeah, way to fracking complicated. Much better to not bother trying to make finding the specific category you want easier, and just go with silly solutions that waist people's time, because they can be gamed, and don't improve the search system at all. The only halfway good idea I have seen so far is Amity Slade's, and imho, this might even be a useful addition to my ideas, since it would increase rank "within the existing search", for places that sell specific types of items, while limiting word salad. It however does nothing to address the issue of phone book style lookup, where you want to know where every kiosk for Luskwood Creatures is, or where the main store is located. The two are tangential to each other, but both imho, more valuable than trying to treat SL like its indexing fracking web pages. There is a reason why phone companies don't bloody use Google to look for phone number and addresses for businesses, instead of printing copies of local phone books. We are dealing with a world that needs a phone book, but also needs a more general search, and where one/both can "narrow" the focus enough that the other becomes useful, instead of remaining totally useless.

No.

I really seriously doubt that's the case, Phil. My picks consist of my own parcels to make them easier for people to find, and my friends' places. Now I happen to think that my friends' places are cool and worth visiting, but that isn't why they're there. I'm willing to bet that the vast majority of people use them exactly the same way I do.

Surely then, the Picks you have for your friends places are genuine 'votes' for those places. They are statements of "I think they are cool places".

The wish to included metrics is probably just a legacy of the "community" thing.

Thinking about all of this, and of LL's issues with development, database and network resources, here is a suggestion:

Every parcel should have an option to generate a web page from a template.
The pages should be viewable to the Internet at large and hosted under a SL domain name variant dedicated to parcel pages so as to allow, e.g. a Google search within that domain.

Sections of the web page should be editable via web forms by the parcel owner(s).
There could be a standard structure of tagged sections for products, events, etc.
People could add pretty unlimited/generous amounts of free text and images.

Other sections would be purely auto-generated from existing About Land data


In-world searches could be run by filling in a form that would generate an 'Advanced' syntax Google search of the Parcel domain and pass this to the embedded or external browser.


The benefits would be that apart from the page-generation software, all of the hosting and searching is bog standard web stuff. SL could grow to a huge extent without any scalability issues for Search.

Why mess around with custom search when most of what is required already exists and in a format familiar to all Web users?


Another benefit for SL would be that many normal RL web searches would begin to pull in SL Parcel pages because of the free-text content.
It becomes viral marketing for SL.
A benefit of that for parcel owners could be that the backlink from the parcel pages to the SL website could contain the Refer-a-Friend reference, enabling a bonus to be paid to parcels owners whose pages draw in signups.

And what if it's the worst place in SL for what it does? People aren't adding it to their picks because they care, but because they get paid for it. What do they care if it's a good or bad place?

Places that currently game traffic aren't shiny beacons of "SL's finest" or noone would be complaining that traffic needs reworking other than store owners.

Unless you can show that the majority of people currently influencing their ranking in Search / Places aren't really doing any harm because they're the "best places in SL" you will have those same "worst places in SL" shift to influence their ranking in the new search as well.

If you deliberately influence your ranking upwards you are making results less revelant because *you* are not in an objective position to decide whether *your* store should rank above your competitors. All you're doing is shifting relevancy for a keyword to something that better agrees with your personal opinion, you have no basis to claim that everyone else will see it the same way.

I supposed they could check if the user is running the offical client and only count those that are, but that would be a temporary bandaid at best since LL wants there to be a lot of legitimate 3rd party viewers.

Bots can be made to do all of these things, and they would if there was something looking for that kind of activity. Even the "anti-bot" camping gimmicks have workarounds, bot runners haven't jumped on them much only because there are easier pickings right now.

Yeah. Horrible idea. Especially since about the only way the non-premium people can get any money is by camping some place. There isn't really any way you could "test" them to determine if they where not bots, so you could mark them as "Ok, track this avatar." First issue is obviously, who do you have do that? Second is, how do you keep them from marking their own bots are legit? Nope, bad idea all around.

It sounds reasonable, but I do see two downsides:-

(1) When people inside SL do searches, they don't want stuff that's external to SL in the results, and I'm not sure that an Google Advanced search would be useful for ensuring that.

(2) It would attract the same objections that everything else has attracted - people are able to influence the rankings. They would do that by optimising the page content. I was an SEO until about a year ago, so I'd have a good start on that, and I'm not the only one who could do it right off. We might even see SEO services in SL, and then there'd be some shouting

You can, bots (and campers for that matter) will have a usage pattern that stands out from an interactive user's.

The problem is that if it's known that "activity" counts, you're entering a race with bot owners constantly changing their bots to try and more closely match a human's usage pattern.

That still wouldn't be a reason not to try it if only for the fact that LL's tedency is to simply throw in the towel (first land was abused with alts, instead of trying to remedy that LL just got rid of it) or state that it's not interested in even trying to compete in a race they'll never truly win (the copybot/open source statement).

If you haven't followed it, you might want to read through the "Traffic Alternatives" thread (/327/c8/256382/1.html). My proposal would simply make camping bots irrelevant, keeping LL from having to do anything about them directly. The remaining L$ camping farming bots would stick around, but those would actually help undermine the effectiveness of camping so you turn the tables there .

Thought this was interesting, then re-read it. lol It wouldn't fix anything, since you are still stuck with a) word salad, if you want someone to be able to "describe" what is there, and b) it would generate the same mess we already have. One of the things I considered, but rejected, was region search. The reason being that you can't even "find" 100% of all regions on a map, never mind search for them, if you don't know their names. Now, one *might* set it up so you can see what region something is in, then click, "Show other things in this region.", sort of like, "Show other things in this mall.", would work for those that designated themselves as shopping malls. But that is about it. Otherwise, if you can't "find" every single region in SL, and half of them are literally not connected to each other, how the heck do you find parcels in any meaningful way?

Oy, I'm not sure about that. Most of the Picks that I see are pictures of friends, lovers, etc, and have no correlation to the parcel on which they were added.

I wish I truly understood how search engines work. I swear I've read the explanations over and over, and they're still a mystery. I do, however, wish that LL would just eliminate all traffic and all "popularity" issues (i.e. Picks) just because it all seems like such a freakin hassle. I'm all for the alphabetical Places and keyword-run All tabs.

(1) That's why all the parcel pages would be under a dedicated domain.
The Advanced Searches would begin with e.g. "siteecondLifeParcels.com"

(2) The SEO options would mainly be down to the user's own text content. The pages would be generated via forms.
If people's words can influence the Google rankings, then good luck to them - because that's exactly how it should be

SL already indexes any object on the parcel that's set to show in search, so if someone searches for "female skins" it should look at keywords in parcel descriptions, then at the contents of the parcel. Between two parcels that both have female skins in their description, the one that has the most objects marked for sale with names that contain some or all of the keywords should score higher.

Of course then people would just put bogus objects for sale buried deep in their basement where no one would ever see them.

Word salad is good.
Let people describe what is (purportedly) on their parcel.
The "mess" that we are concerned with in this thread is that caused by the subversion of traffic and other metrics by people gaming the system.
If the only "gaming" open to people in general is that of composing text for their parcels then that's wonderful.

I wouldn't be too concerned about regional/area searching in a TP world.
A quickie way of facilitating this would be to have each parcel web page include a grid of icons for the 9 adjoining regions centred on the parcel region.
The icons could be taken from the existing Map.
Click on any of the regions and you'd get a listing of all parcels in that region with summary details of each.

I'd like to be able to have both a word salad and an area for a decent, well-written description. Have the search engine use only the word salad for its results; leave the description out of it altogether.

The problem there is that in my suggestion we'd be using standard advanced Google search, and therefore limited to the the options afforded by that.
Of course, if/when Google improves the way in which searches can be put, then SL and the users all get a freebie benefit.

Even if using standard web searching techniques (albeit a bit assisted by an in-world front end) might seem a bit clunky, there are huge advantages in using it.
It takes a huge current and constantly growing load away from LL's servers, network and development teams.
That's *GOTTA* be good