Deciphering Matt Cutts Options

[laura]

From Stephen Spencer blogging Matt Cutts' words at the SES conference:

1. Clicks on Nodes:

"In graph theory, a clique in every node in the graph is very unnatural. So don’t link to every single node in your network of sites; it’ll get flagged." 

a click on every node? Dont link to every node? Does this mean, dont link to every page in your network of sites from all of the other pages in your network of sites? (Eek. Scary.)


2. Google dropping parameters

"For dynamic sites, you’re very safe if you have fewer than 2 parameters; keep the values of those parameters to fewer than 5 digits, and don’t name a parameter “id”. Googlebot sometimes tries variations of URLs by dropping parameters, but we only do that deep level analysis on big, quality sites."

The parameters bit - understandable. But Googlebot tries variations by dropping parameters? (IMG:http://www.highrankings.com/forum/style_emoticons/default/eek.gif) What does this mean, really?

[clueless]

I am going to piggyback here because I've been wading through the same stuff. I am concerned about the id parameter. I have them. G indexes them. So what does that mean? It seems that a glimpse behind the curtain raises as many questions as it answers.

Thanks Laura for bringing it up - it's nice to know other people are spending time confounded by this stuff.

[Jill]

Can you guys provide a link to the original article? Sometimes it helps if we read the entire article in context, rather than just the snippets. (Plus we should link anyway when something is quoted.)

Thanks!

[laura]

The article on SES link Q&A session.

[Jill]

a click on every node? Dont link to every node? Does this mean, dont link to every page in your network of sites from all of the other pages in your network of sites? (Eek. Scary.)

Yes, I believe that's what that means. I didn't attend the session which Stephan wrote about, so I can't be sure, but that's how I read it.

I'm not sure about the other one. I think it should say don't use &id not just "id" though. And by dropping parameters, I'm guessing that perhaps Google will try to cut off the ends of urls at a question mark or other query string in an attempt to get to the root url. This would stop them from indexing content under various URL strings. (But I'm just guessing.)

[DanThies]

Googlebot tries variations by dropping parameters?  (IMG:http://www.highrankings.com/forum/style_emoticons/default/eek.gif) What does this mean, really?

For some (important) sites... if they find a URL with something that looks like a session ID variable, they might try fetching the URL without that variable, to see if they get the same content, and can safely surf the site without that variable.

Does this mean, dont link to every page in your network of sites from all of the other pages in your network of sites?

Sure, although he's probably talking about the more usual scenario, where people have multiple sites and run sitewide footers cross-linking them all. Linking to the home page of every site you own from every page of every site looks really unnatural.

You can run sitewide links with rel=nofollow and that oughta be safe, or use a redirect (and robots.txt) to keep the spiders from following the links.

[Ron Carnell]

For dynamic sites, you’re very safe if you have fewer than 2 parameters ...

Sigh. Google does it again.

So, did Matt mean what he actually said? Or did he fail to say what he really meant? (IMG:http://www.highrankings.com/forum/style_emoticons/default/sad.gif)

[Michael Martinez]

A clique would be more like a link circle, in which every page links to every other page, rather than every page simply linking to all the home pages.

For example, if you have a network of 3 domains, each with 5 pages, to form a clique, you would have to have include 14 outbound links on each page to all the other pages in the circle. The distinction between the domains would be lost, in terms of natural linking relationships.

This is a legitimate concern, in my opinion.

Most of my domains do actually link back to the home page of my primary domain. The backlinks are on most but maybe not all of the pages. In that respect, since my primary domain pages don't link back to all the other pages (I cannot imagine putting 50-100,000 links on a page, to be honest), I feel pretty safe that I have not formed a clique or anything resembling that.

[laura]

Thanks all, for putting this in more digestable terms (IMG:http://www.highrankings.com/forum/style_emoticons/default/flowers.gif)

We have more than 5 digits in our parameters, but we strip the tags for search engines anyway (in most cases). Amazon's URLs can be pretty enormous. So I'm not too worried about that I guess.

[stephan]

Since I wrote the blog post that we're all dissecting, I thought I should probably jump in to clarify.

Cliques

First off, it's a clique not a click. I don't want to bore you all with Graph Theory (although if you're interested, you could read up on it on Wikipedia at http://en.wikipedia.org/wiki/Graph_theory ), so I'll cut to the chase. Simply look at the figure on the page http://en.wikipedia.org/wiki/Clique_(graph_theory) and you'll see that all 5 nodes (these are the signified as red dots and they're referred to as vertices in the caption but a node and a vertex are the same thing) are all "linking" to all 4 of their neighboring nodes. They never miss a link. It all looks so perfect, doesn't it! Naturally occurring neighborhoods on the Web aren't perfect like that. If it looks perfect, it's been engineered. Google will be suspect to unnatural-looking neighborhoods.

Googlebot and parameters

When Matt said that Googlebot sometimes tries variations of URLs by dropping parameters, he meant that Googlebot may experiment with dropping name-value pairs from URLs. I understand the reason for this to be that if the page still shows the same content, it gives Googlebot an indication that the omitted variables are superfluous in the query string. So for example, a URL such as this:

www.bigyellow.com/cgi-bin/php/cities/unitedstates/mtg_detail.php?SRC=&PID=36575&S=NY&T=&MTG=PR

might be shortened by Googlebot to:

www.bigyellow.com/cgi-bin/php/cities/unitedstates/mtg_detail.php?SRC=&PID=36575&S=NY&T=

and

www.bigyellow.com/cgi-bin/php/cities/unitedstates/mtg_detail.php?SRC=&PID=36575

and

www.bigyellow.com/cgi-bin/php/cities/unitedstates/mtg_detail.php?S=NY&T=&MTG=PR

etc.

Then these URL variations would get spidered and compared with each other. I've heard of big websites getting hit by this and it causing big problems for the website in question.

Regarding using id as a variable name... Matt advised webmasters against it because Google is suspicious of that variable being a session ID or something other than a key field. Even if it's the only variable. Particularly if that variable's value is long. sid would be a bad choice too because it could stand for session ID as much as it could stand for a key field like story ID. It doesn't mean that your pages won't be indexed if you use this variable name; it just means those pages would be at a greater risk of not being included. You should be fine though if your pages are all in Google.

Hope this helps.

[Randy]

Welcome Stephan ! (IMG:http://www.highrankings.com/forum/style_emoticons/default/hi.gif)

Thanks for the clarification.

[Jill]

Welcome stephan! (IMG:http://www.highrankings.com/forum/style_emoticons/default/bye1.gif)

Nice to see you here, and nice to have met you in San Jose! (IMG:http://www.highrankings.com/forum/style_emoticons/default/smile.gif)

[stephan]

Hi back, Jill and Randy! Thanks for the warm welcome!

I decided to turn my clarifications into blog posts:

www.stephanspencer.com/archives/2005/08/26/how-graph-theory-relates/

www.stephanspencer.com/archives/2005/08/27/googlebot-parameters-and-dynamic-sites/

If there's anything else from my original Matt Cutts post that you would like further clarified / expounded upon, just give me a yell and I'll be happy to do so. (IMG:http://www.highrankings.com/forum/style_emoticons/default/smile.gif)


<live links removed>

This post has been edited by chrishirst: Aug 28 2005, 04:03 AM

[storyspinner]

(IMG:http://www.highrankings.com/forum/style_emoticons/default/bye1.gif) Hi Stephan!

Thanks for the post clarifying all that. This was really packed with some good info.

That was a session I missed, but I did attend one of yours! And actually read your blog pretty regularly too (IMG:http://www.highrankings.com/forum/style_emoticons/default/smile.gif)

[ewc21]

Welcome Stephan! (IMG:http://www.highrankings.com/forum/style_emoticons/default/smile.gif)

And thanks for your blog.