BigDave - 6:59 pm on Mar 8, 2003 (gmt 0)

GoogleGuy,

I really appreciate the additional communication that we are seeing lately. It seems like someone at the old googleplex has decided that it's okay to give out a bit more information about how things operate, as long as they do not negatively influence the integrity of the algo.

Having worked on many quality software teams, and too many that weren't, I have no trouble believeing that you all are able to cover all the common cases before implementing. I'm certain within a few months of bug reports, you will have this particular piece of the algo almost perfect.

I have a couple of suggestions that you might want to consider. You may have already implemented them, or you may have already decided against them. But as free suggestions, they are worth every penny you paid.

The suggestions mostly have to do with the problem of someone buying an expired domain widget.com with the legitimate goal of selling widgets. They will have trouble getting "new" links from many of they places that would link to the site, because those sites "already link to them".

One possibility is to reinstate the link after a certain amount of time. Possibly something like letting them start trickling back in after 6 months, lowest PR links first. This will give the linking site owners more time to realize that there has been a change.

Put together a filter that compares the words on the old site with the new site and come up with a probability that it is on a similar topic. This one could be done, but I'm not sure if it's worth the effort unless you already have someone putting effort into any sort of baysean analysis for themeing.

When a linking page has had some significant changes, specifically, having several links removed, then allow that link to count again. That way, if I am linking to widget.com and if I go through and check my links to delete broken ones, I am sort of validating that I still intend to give that link to widget.com. On the other hand, a list of links that only gets added to might never get verified.

Sometime it might pay for google to document the process that they go through when implementing one of the more simple spam filters, so that people can see the process that goes into making sure that you do not catch the dolphins. Certainly not one of the major spam filters, and you don't need to give all the information, but enough so that we can point to it and show people that you really do try to avoid killing innocent bystanders.

On another note, my biggest wish for google by next fall, would be for Google to either have a way for webmasters to request that they update their entry in their DNS cache, or tell webmasters how to have their hosts transfer the DNS so that google will catch it. I'm hoping to exceed my bandwith limit in about 6 months, and it would be nice to not have to keep my old site up for 3 months just to be safe.

Thanks again,
Dave