Yeah, for something like a blog I would assume you would link not to the front page but the page for the entry or time period (depending on the blog), at which point this solution seems eminently reasonable. If anyone does do a Firefox extension, it'd be fantastic if they'd be kind enough to drop a link here.

Re: JavaScript, I was thinking mainly of embedding something that would catch the #=.* and highlight/jump to the appropriate text. Check document.location.hash, I suppose? If support was offered in a few widely used libraries, it might be a wedge for creating a de-facto standard. Still not as immediate an element of the UI as some kind of direct hook to the browser's builtin search, of course.

But a Firefox extension would work great as a proof-of-concept demo...

Linking into a dynamic page is not necessarily going to be reliable, but then linking to a dynamic page is not necessarily going to continue to mean anything consistent, either, so this is not really an objection. Doing the reference with an XPath would basically miss the point, since you don't want to be pointing to "whatever happens to be in the third paragraph now", you want to be pointing to the text that was the third paragraph when you pointed to it. So the search string seems to me to be the more reliable method. And implementing it by literally triggering the browser's text-search function means that if the offset is no longer right, due to page changes, the user can use the Next/Previous buttons to see the other uses of the phrase in question.  

There's no cross-browser Javascript method for bringing up the browser's Find mechanism, and if there was, where would you put your Javascript code?! The idea is only practical, I think, as an addition to the standard. People won't use it unless they can count on browsers supporting it...

I happened across the Church of Purple today, who look like they're interested in the same sort of thing. Referenced from this code, which implements an xpath-based syntax that would be very useful for Glenn's idea.

This seems like a reasonable method for indicating search strings within the text at a given URL, and I suspect it might be reasonable to implement in JavaScript (though I don't know of a mechanism offhand). On the other hand, I think there's a basic semantic problem with using the the search-string + offset method to refer to a particular piece of text: Few, if any, URLs are guaranteed to represent a static piece of text. This falls down, for example, on the front page of most blogs or any given wiki.

I am barely competent in HTML tags, but glenn's searchable link solution here is 1. brilliant and 2. the answer to something I've wanted to be able to do for a long time. Can someone or has someone brew up a widget for Firefox that does this? I'd do it myself, but I suffer from a complete lack of knowhow.