Some posts and advice on Bourbon and other topics
| 3:12 am on Jun 2, 2005 (gmt 0)|
Hey moderators, let me know if there's any way to lock this thread for a while so that only I can post in it. That would let me put out a few posts that won't get lost in the Bourbon shuffle. I think maybe WebGuerrilla set it up last time?
| 5:57 am on Jun 2, 2005 (gmt 0)|
Go ahead in this thread Googleguy.
Fellow WebmasterWorld members - this is Googleguy only for the moment - all other replies will be deleted for the moment.
| 7:52 am on Jun 2, 2005 (gmt 0)|
Many thanks, vitaplease and Woz and the other mods who keep things running so well here. I've got to run a meeting early tomorrow morning, so let me just start by noting a few things. I'll start by repeating the message #567 from thread 3. reseller was asking for a little bit of an indexing outlook for the next few day and how much flux to expect. I said:
|reseller, currently, a few data centers have some different data that should be everywhere in a few days. I'll keep people posted on the status of things, and collect feedback closer to the end after things are settling down more. I'd expect that things will be back to their normal level of everflux by New Orleans. But we do have incremental indexing after all, so it's normal to expect a certain amount of change to the index every day or so (aka everflux). |
In fact, everflux is a pretty good analogy. If you go back to summer 2003, update Fritz was the beginning of the transition from a monthly update to an incremental index. It caused a lot of comments, because plenty of people were happy with an index that only changed once a month. A lot of the thickness in my hide started with Fritz during summer two years ago. happy! Summer in the northern hemisphere is often a good time for a search engines to work on revamping different parts of our system and improving our quality; typically search engine traffic is lower in the summer due to seasonality. So the summer is a good time to think about things like bringing in new signals of quality and ways to rank pages, plus doing things like reorganizing our webmaster pages, etc. etc.
It's true that the summer (northern hemisphere) is when traffic is lower and sometimes it's easier to roll new things into crawling/indexing/scoring. I wouldn't be at all surprised if we worked on revamping our webmaster pages, for example. The SEO and quality guidelines pages have aged pretty well, but other parts of the webmaster section need to be reorganized; there's a few places where there's the same info (e.g. about robots.txt) repeated in several places or scattered over different pages. It's not trivial to reorganize that much info, esp. since it's translated into so many languages. But better to go ahead and start, and then if we want to tune those pages later, that's okay. I've been aching for a long time to mention somewhere official that sites shouldn't use "&id=" as a parameter if they want maximal Googlebot crawlage, for example. So many sites use "&id=" with session IDs that Googlebot usually avoids urls with that parameter, but we've only mentioned that here and on a few other places. Getting started on things like that will be nice. I appreciate the people who sat down and tried to tease out the info on our current webmaster pages and organize it more logically.
| 7:59 am on Jun 2, 2005 (gmt 0)|
taps, in message 33 of the first Bourbon thread you mention that you sometimes see your meta description shown instead of the typical Google snippet. That's something that we've been experimenting with for a little while. To be honest, I'm not sure what the exact criteria are for using the meta description instead of a normal snippet, but it's independent of the update. You can assume that we're always looking at different ways of giving snippets (from the text of the page, from an ODP description, from meta descriptions) and trying to find what's the most useful result for a searcher.
I have to admit that I'm so conditioned to our regular snippets that when we show one of the other types of descriptions, it's a little jarring for me. But I've been keeping an open mind and trying to figure out when I like it vs. when it just keeps me from getting the info I want faster. :)
| 8:07 am on Jun 2, 2005 (gmt 0)|
One thing that I noticed in the first 300 posts or so was a discussion about whether backlinks or PageRanks were also being updated at the same time. In general, I wouldn't fixate too much on backlinks or PageRank during an update. The external backlinks/PR that we show have at least a couple factors that complicate analysis. First off, they are a snapshot in time, and the actual backlinks/PageRank used for ranking are from a different time interval. Another complicating factor is that we don't show every backlink on Google; we only show a subset of the backlinks we have internally. [One of the things we needed to do on our webmaster pages revamp was to go back and make sure it's clear that we only show a sampling of the backlinks we know about, not all backlinks.] The snapshotting + asynchronous nature + subsampling actions means that it's pretty hard to trace the reason for a ranking change back to a particular exported view of backlinks or PageRank. This is an area where I wouldn't read as much into the particular backlinks or PageRank that you see at any particular instant in time. Also, if a backlink or PageRank external update happened at the same time as ranking changes, I wouldn't read too much into how closely in time they happen together.
Just as a guide for people who don't eat and breathe WebmasterWorld, we typically show new backlink sample sets every 3-5 weeks or so. We have a bank of machines that computes PageRank continuously (and continually, I suppose; I wasn't an English major), but we only export new visible PageRanks every 3-4 months or so.
| 8:15 am on Jun 2, 2005 (gmt 0)|
One thing that is more clear if you read update threads in one big chunk (instead of in-the-trenches as posts are being made) is how folks can get off onto different (and sometimes mistaken) tangents. One of those tangents was that Google would penalize pages that had <iframes>.
I've never heard the suggestion that Google would penalize for iframes before reading it in the thread. Plenty of legit sites use iframes, so it wouldn't make sense to penalize for it. Powdork gave the right response in message 337 of the first Bourbon thread. Now I can easily believe that some search engine spiders would have trouble with iframes just like some spiders have trouble with frames. But I wouldn't expect iframes to cause any penalties.
| 8:26 am on Jun 2, 2005 (gmt 0)|
Sometimes a tangent isn't such a bad thing though. For example, partway into the Bourbon discussion, wattsnew asked if there was a technical guide on how to handle www vs. non-www, relative vs. absolute linking, and links to different pages such as / vs. index.html vs. default.asp. My rule of thumb is to pick a root page and be as consistent as possible. I lean toward choosing [yourdomain.com...] but that's just me; [yourdomain.com...] would work as well. Then I recommend that you make things as simple as possible for spiders. I recommend absolute links instead of relative links, because there's less chance for a spider (not just Google, but any spider) to get confused. In the same fashion, I would try to be consistent on your internal linking. Once you've picked a root page and decided on www vs. non-www, make sure that all your links follow the same convention and point to the root page that you picked. Also, I would use a 301 redirect or rewrite so that your root page doesn't appear twice. For example, if you select [yourdomain.com...] as your root page, then if a spider tries to fetch [yourdomain.com...] (without the www), your web server should do a permanent (301) redirect to your root page at [yourdomain.com...]
So the high-order bits to bear in mind are
- make it as easy as possible for search engines and spiders; save calculation by giving absolute instead of relative links.
- be consistent. Make a decision on www vs. non-www and follow the same convention consistently for all the links on your site. Use permanent redirects to keep spiders fetching the correct page.
Those rules of thumb will serve you well no matter what with every search engine, not just with Google. Of course, the vast majority of the time a search engine will handle a situation correctly, but anything that you can do to reduce the chance of a problem is a good idea. If you don't see any problems with your existing site, I wouldn't bother going back and changing or rewriting links. But it's something good to bear in mind when making new sites, for example.
| 8:39 am on Jun 2, 2005 (gmt 0)|
Okay, now I'm all carpal tunnely from typing on the laptop. Here's a recommendation that I hope everyone at least knows about. On WebmasterWorld, if you click on "control panel" at the top of the page, you can select to show up to 30 messages on each page of a thread. Highly recommended for long threads.
| 6:50 pm on Jun 2, 2005 (gmt 0)|
I'd recommend doing it for any link on a site, if you have the ability. If the amount of work is high and things seem to be working well already, I wouldn't worry about it though.
| 6:56 pm on Jun 2, 2005 (gmt 0)|
Hey, I talked to someone who was familiar with our re-org of the webmaster pages, and it sounds like it's been going better/faster than expected. I said we'd probably revamp them this summer; it looks like we'll be likely to put a first new version of those pages out quite soon. I expect we may iterate a little more if there's info we can add in there.
Like I mentioned before, summer is a good time to work on pulling in new pieces of infrastructure and ways to rank/score pages. I'd look for us to keep finding new ways to get info out to webmasters (having my own roomy thread to post in helps! :), and hopefully provide more ways for webmasters to give info back to us as well. :)
| 12:17 am on Jun 7, 2005 (gmt 0)|
Okay, a bunch of reinclusions (and the responses to some spam complaints) went out today, along with some improvements to one of the first two things that went out with Bourbon. I also believe that we completed a binary push this weekend that has hooks for the last part of Bourbon. We'll turn on one datacenter, and then the rest of the datacenters over the course of the next week or so. After the other changes that went out, this last change should be less noticeable. In a few days, I'll post a way to give feedback, but feel free to use the the spam report form at [google.com...] in the mean time if you want to mention spam that you see in the results. I'd use the keyword "bourbon" if you do decide to use that form, so that we can separate those reports out.
| 8:54 pm on Jun 10, 2005 (gmt 0)|
I just found out something about reinclusion requests. If you send in a reinclusion request, you'll get an automatic reply with a tracking number in the subject. For best results, if you need to discuss that site in the future, it's best to reply to that email instead of mailing with new messages. Replying to the message keeps the same tracking number, so that the person who evaluates the site for reinclusion will get the complete correspondence about your site. If you send lots of disconnected emails so that you've got lots of different tracking numbers, the person who is evaluating the site for reinclusion won't have the fullest picture available to them. Just a tip for anyone who is doing a reinclusion request.
| 12:22 am on Jun 21, 2005 (gmt 0)|
Where was I? Ah, I just saw that Australian users aren't seeing PageRank. Do not panic! We're just testing something. You'll probably see PageRank again in a while after we're done testing..