My browsers haven't yet got the hang of caching data. Now google's doing it too?

are the rules for elements always the same no matter where in the structure they appear?

Using DTDs, yes. If your validation tools support it, there's an alternative to DTDs, called RelaxNG, that can do what you want.

Dang. Thanks, Tom. I've been arguing this with my co-worker for like ever. It seems weird to me that you can have all of this structure but not be able to use it to change the rules for elements within them. We use XMetaL as our editor and I know you can "hide" elements there that are still in the DTD so that's probably what we'll do.

It seems weird to me that you can have all of this structure but not be able to use it to change the rules for elements within them.

That's why RelaxNG was invented.

And so they're hitting the server whether you follow a link or not.

It's the other way around. If you follow a link, they save it. The next time you follow the link, they give you the saved copy unless there's reason to believe it's out of date. So you wind up with fewer server hits, not more.

Now google's doing it too?

Forget it, Jake, it's betatown.

There definitely used to be web accelerators that ran on that principle. They'd grab the text of links, then the images, using a choice system based on your habits. They were actually quite good (though, yes, meanish to servers) when reading text-heavy pages with fewish links. Now that the internet tends to have billions of links per page (as it should, but didn't for a long time), they probably don't work quite as well.

So will my referrer logs stay updated properly? How does that even work with browser caching?

I can't answer the detailed questions; (A) not my project and (B) not authorized to speak for the company. Everything I know is in the FAQ.

[ita: you probably want the Webmaster faq. [link] ]

Reading the FAQ leaves me with the same impression I had before. I said:

Without knowing what I'm talking about, I got the impression from signal v. noise that google web accelerator pre-caches all links when you're on a website. And so they're hitting the server whether you follow a link or not. Does this sound like I'm onto it?

And Betsy said:

It's the other way around. If you follow a link, they save it. The next time you follow the link, they give you the saved copy unless there's reason to believe it's out of date. So you wind up with fewer server hits, not more.

But this sounds like regular browser caching. If google says below

What is prefetching?

Prefetching takes advantage of idle time when a user is viewing a page to accelerate the links which he or she is likely to follow next by “prefetching” those pages. This speeds up users' online experience by making many pages load faster. You can learn more about prefetching on the Mozilla website.

2. How does Google Web Accelerator decide what to prefetch?

Google Web Accelerator decides what links should be prefetched based on aggregate usage statistics as well as the user’s mouse movements. But Google Web Accelerator also allows websites to specify which links should be prefetched (for example, the top story on a news website), so that webmasters can prioritize links which they expect their users to be interested in.

I still got the impression that it's pre-caching, and the signal-to-noise people think this will increase load because you won't necessarily follow that link.