What is the Poodle vulnerability in SSL and what should you do about it?
SSL works with a number of different mechanisms. One of those is a creaky old thing called SSL 3.0. Most Internet browsers and web servers still support SSL 3.0, but researchers have just discovered a vulnerability in this one protocol, whereby if someone can insert a computer between the computer initiating an SSL connection to a server (e.g. a user logging into a website over HTTPS) and the server itself - at any hop in the chain - they can potentially eavesdrop on any information sent in that supposedly secure channel. That's called a Man In The Middle attack, or MITM for short, and the basic principles are described quite well on WikiPedia.
There are a load of ways people can instigate a MITM attack, for example you might be in a café using a shared wireless network and not realise someone else is pretending to be the Internet router, so all your traffic is channeling through their laptop before heading out to the wider world. It's surprisingly easy.
So assuming this is worrying enough and you don't want to simply accept the risk exists and ignore it, you'll want some kind of mitigation plan. The question is what are the paths avaliable for mitigation and which one should you take?
You could take the view you can deal with this just with policy and procedural change. For example, if you decide most traffic encrypted over SSL is non-critical, such as end user logins but users with no real power to break anything or carry out malicious activity, then you could alter procedure for users with elevated levels of access, insisting they only login for certain known networks, where a MITM attack would be practically impossible to mount, such as only using a physical cable and within the confines of your office network. Of course, you cannot enforce this (though you could be notified of breaches using realtime log inspection tools like OSSEC) but at least you can educate more privileged users and significantly reduce the risk of the possibility of the conditions existing in which a MITM attack could more easily be carried out. And you'll also be making the rather grand assumption no one is doing anything they shouldn't be doing in your internal networks (or you're well set up enough to know if they are and catch them).
However, our preferred approach - the one we're recommending to our hosting customers, and indeed we do ourselves, hence our A+ rating at SSL Labs (hat tip to Miguel Jacq) - is to simply not accept SSL 3.0 connections any more. Happily, as I mentioned earlier, SSL provides several different methods for encrypting network traffic and SSL 3.0 is just one of them. You can just switch it off!
There is a but though. If you do that, you can no longer support Microsoft Internet Explorer 6, because it is simply too old to work with the more modern options. In most cases you shouldn't care. According to Microsoft themselves, IE6 is now (often significantly) less than 1% of the browser market in Europe and North America, as well as large swathes of the rest of the world. If you have many end users in China or Africa who use your site over HTTPS you might want to think on this though, as in some countries IE6 still hovers around 10%, sometimes higher.
We believe education is the way forward, all of these people in countries where IE6 usage is still high have the ability to download and use a free, secure alternative, such as Google Chrome. We prefer to tell people to stop using IE6 over perpetuating insecurity on the Internet, but that's our choice. You are perfectly entitled to disagree with us, of course.
In any case, we hope you found that article useful and it clarified your position with regard to this latest SSL vulnerability.