At the Cloud Identity Summit last week, Craig Burton stated the SAML – the security assertion language that helps thousands of enterprises address single sign-on – is unequivocably dead. Kaput. He presented the following data points to support his argument (I will link to his presentation when available):
- Proliferation of APIs: There are so many APIs, billions in fact, and we have thousands popping up every second, that we cannot ever hope to integrate them with SAML. The effort is too great, and integration is too complex, for all the services to address the scope of the problem.
- Scalability: SAML cannot scale to solve the cloud’s many-to-many problems, and is too cumbersome to address such a large problem.
- Lack of support: His final point is that all the major backers have stopped financial support for SAML 2.0, and it appears that no one is driving advancement of the standard. Without more support fundamental limitations in the standard simply cannot be addressed, and support is shifting to OpenID Connect.
Three solid points. But do they mean SAML is dead? And what the heck does ‘dead’ mean for a product anyway?
One the first point, I disagree. There are different ways to scale solutions like SAML, and there are indeed billions of APIs, but we do not we want or need SAML to give us SSO to all of – or even a significant fraction – of them. That’s a rather silly utopian dream. And the lack of support for revising a standard does not mean that it is obsolete – or that we should stop using it. That’s my take and I’m sticking with it.
I was originally going to title this post is “SAML IS Dead”, but that’s not what we should be talking about. SAML’s longevity, and how much faith customers should put into technologies called ‘dead’, are only part of the problem. This most recent claim is only one instance in a long-running series. We have seen people – no, let’s call this one correctly – analysts – say stuff is dead. All the freakin’ time. IDS, anyone? How many people have said Windows is dead? It’s like any limerick that starts out “There once was a man from Nantucket …” – after the first few you know the pattern.
To an analyst there is value to doing this. Advising customers when a technology has been superseded, or will likely be obsolete within a few years, is useful. It helps companies avoid selection of suboptimal technologies, and investment in inferior choices when better options exist. But labeling something ‘dead’ has become every analyst’s favorite way to be a drama queen. It’s to get attention, and to exaggerate a point when you don’t think your audience is paying attention. I understand why it happens, but it’s not helpful. It fails to capture the essence of the slow evolutionary replacement of technologies. History has shown it’s just as likely to be wrong, to mislead customers, or both.
Why call products dead when everyone is still using them?
Many people on Twitter had the same thought I did – PKI, IPV4, Kerberos, AV and firewalls, have all been ‘dead’ for years – but they all remain in wide use with no indication of actually going away. Worse, when we say older standards are now made obsolete by new standards – which are yet to be finished, much less adopted – we often fall on our faces when the new standard gets stuck in committee and turns out to die while the ‘dead’ predecessor lives on. We have seen cases where simplicity of concept (UNIX) trumps a grand vision (MULTICS). And we have seen cases where technologists want something to die (IE 6 comes to mind), but the general population sees value and utility in the product. Plenty of technologies which are wished or “supposed to be” dead continue to be essential computing and security. So maybe dead means “dead to me” – an entirely different meaning.