2by4 - 9:19 pm on Oct 3, 2005 (gmt 0)

If you accept this idea is possible, then the topic of this thread becomes much more understandable. If it's a new system, then there are going to be issues which need to be fine tuned since all beta software has issues. The way google handled duplicate content previously was as far as I'm concerned absolutely primitive, pathetic really.

I've suspected that google has switched for a while, but caveman's observations really clicked something in my head, I've been so busy saving client's sites this summer that I hadn't had time to really step back for that overview of just what it was that lay behind all the stuff I was doing, but now it's starting to click:

Everything I did this summer to save sites that fell, and am now having to always do on every site I do, involves a significant tightening of discipline, no errors allowed. Previously we got away with massive errors. What does not support this type of error? A brand new system written to be tighter, that tightness gives greater control, but also hurts people who were able to get by with errors in the past.

All sites I do that were written from the beginning with no errors of this type, or where I fixed these types of errors a year ago, have seen zero negative affect in the serps, in fact, almost all are performing better and better. But older sites, with sloppy components, code, webmastering, dupe content, etc, are behaving RADICALLY unpredictably. Huge fluctuations, corrections, new fluctuations.

In other words, when I feed google clean, unambigous data, it works very very well. When that data is corrupted, it becomes extremely unstable internally in the google algo. This is exactly what I would expect from a new algo, since it's exactly what I see with my own web programming with each new thing I develop.

Just like say using XHTML instead of slop html, or switching to objects instead of running all spagetti code [say windows nt 5 versus windows 98]

If this is a rewritten algo, obviously the google engineers would have rewritten one of its weakest components, the dupe filter. But a new piece of software has to be tested, then you have to start pouring real data through it. You can only test in the labs so much.

I don't believe google had the luxury of postponing this rewrite like say Microsoft has with Vista, I think they were up against a wall and had to do the rewrite, this was becoming fairly evident a while ago as far as I'm concerned.

The relative smoothness of the transition is a testimony to just how good the engineers they've been stockpiling over the last year are. But at some point if you see enough trees it might pay to ask if you are in fact in a forest.

I'm usually happy to criticize google, but since I do enough programming to have some sense of just how difficult it is to introduce new stuff of any comoplexity live, I have to look at the collateral damage slightly differently, you can't come out the door perfect, but at some point you do have to make the decision to walk through it and put it live.

But, again, the dupe content issues caveman raised seem much more like beta issues than mature algo tweaks. Which suggests to me that this stuff will improve over time as they analyze the data they're getting.

A lot of other things are explained to as far as I'm concerned. My guess is that the new algo is much less sloppy than the old one, they have after all learned a lot in the last 7 years, and probably had a very long wish list that simply could not be implemented if they tried to force more and more junk into an essentially obsolete piece of software.