Documentation rot is one of the most annoying things when trying to learn a new technology, and the fault lies mostly with Google. "Definitive" sources sometimes instantly become obsolete when a new version of a technology is released - and sometimes they don't. Sometimes they very clearly indicate which version they apply to, often they don't - if you're new to a technology you're unlikely to be aware of the difference.
The real problem is that "relevance" for technical resources has a very different meaning from "relevance" on the Internet at large, because the content pages and incoming links, by themselves, don't contain all the necessary information to judge whether a page is still relevant. As a result, Google is in some ways less and less useful as a programmer's tool - not because Google is getting worse, but because there are more and more layers of obsolete information out there.
> As a result, Google is in some ways less and less useful as a programmer's tool - not because Google is getting worse, but because there are more and more layers of obsolete information out there.
Its got a name, information white out, and can occur for a number of reasons. The example I use is finding information on Steve Jobs from 2000 - 2005 is nearly impossible on Google because the first 50 - 100 results of the query 'Steve Jobs' is filled with eulogies and references to his death.
But news worthy events are only one form of white out, historical weight is another (as the perl example shows).
Disclaimer I work at blekko.com (a search engine) which allows you to specify things like /date=2005 in the search providing semantic intent. Searching through white-out situations is one of the uses of the slashtag system.
Thanks for the disclaimer. Your post is propaganda.
Finding information on Steve Jobs from 2000 - 2005 is actually pretty easy on Google. Search "steve jobs", then on the left-hand bar, click "custom range," and input 2000 to 2005. Done.
I invite readers to actually try the blekko example to see how incapable it is. Searching with /date=2005 : (a) sorts the results by date, such that you the first few pages are full of irrelevant pages from Dec 31, 2005. And, (b) only returns results from 2005 (it's not clear how to search over a time range).
This is a nice theory, but Google manages to keep step if new material is produced. Gabor Szabo's tutorial series is really very new, but is the fourth result. The problem is that there is not enough up-to-date material connected with the term "perl tutorial".
Yes and no. I obviously don't know that many details about Google's ranking algorithms, but content age and incoming links play a big role. This plays against content freshness and rate at which new incoming links are added.
I suspect Google uses more or less the same algorithm for general content and for tech documentation. The difference is, general content - news, discussions, fiction, etc. - rarely loses relevancy, it can only be supplanted by newer, more relevant content over time.
This happens for documentation, but it takes time - and at any given time you end up with a hodgepodge of content for various versions.
There's also no foolproof way to search for content for a specific version - this would go a long way towards fixing the problem. The closest solution is restricting the search results by date, I often have to do that.
The real problem is that "relevance" for technical resources has a very different meaning from "relevance" on the Internet at large, because the content pages and incoming links, by themselves, don't contain all the necessary information to judge whether a page is still relevant. As a result, Google is in some ways less and less useful as a programmer's tool - not because Google is getting worse, but because there are more and more layers of obsolete information out there.