Identity Services should be like Vitamins, not Crack
OK, so it’s a ridiculous title. But hear me out.
Matt Flynn brought to my attention an article in which Dale Olds talks about the need for hosters (companies that provide the platform on which you deploy your Cloud/SaaS applications) to provide identity services (and as Matt points out, security services in general) as part of their offering.
<Side Note>No, I do not have a vendetta against Novell, though these last few blog posts may make it feel that way. I actually really like the Novell gang – Dale, Ben and Nick Nichols among others – and for the most part completely agree with their views on identity.</Side Note>
Now, I am with Dale for the first half of the article. Developers of these cloud applications just want to focus on the business logic that is at the core of their service, and not have to worry about the plumbing items, which would include identity management. This is fundamental service-oriented security principles at play, and the survey Dale mentions reflects this (I would argue that even the one-third of SaaS vendors that said they want to handle identity themselves are either saying so because they don’t know what’s involved or are just not happy with what they are getting from the platform and embeddable components). A good set of identity services goes a long way in making applications agile and more acceptable/appealing to customers.
But then the article talks about hosters using identity services as a way to make their platform sticky, because if the platform owns the user accounts for the service, then the service will be hooked. I actually envision the opposite of that when I think of identity services in the platform – identity services making it possible for the SaaS vendor to switch between platforms easily. What is being described sounds like an Identity Provider, which is a business service, not a platform service.
What the platform should provide, and what most enterprise customers would want, is an Identity Hub service, as opposed to an Identity Store service. This allows the customer of the SaaS application to plug it into their enterprise identity store (usually a corporate LDAP system, but it could also be their Salesforce user store) and also accept incoming identities over the wire, while still freeing the SaaS vendor from having to manage identities. In this model, the stickiness for the hoster comes not from owning the user accounts, but from the QoS of the identity services they are providing to their customers (the SaaS vendors and their delegated customers). It also doesn’t force a SaaS vendor to be married to one platform.
Now, I am going to be a little presumptuous here. Having spent some time with Dale, and knowing his past work, I think that he believes in the view I am taking as well. The article seems to be discussing the topic of identity services from a particular angle, which is that there is currently a market opportunity for hosters to leverage the lack of good (non-enterprise) Identity Providers to make their platforms more sticky. It is absolutely true that platforms can (and are actively seeking to) make themselves sticky by owning the accounts; Dale points out that this is exactly what Google did by leveraging GMail as the gateway drug (see, I told you the metaphor works). But as Google seeks to penetrate the enterprise market deeper, even they are recognizing the need to support federated identities as a necessary step for viability. (UPDATE: An old blog post of Dale’s actually clarifies this, and in essence agrees with the view point I am stating here – exactly as I thought he would 🙂 )
Bob Blakley has long mused about what business models would make Identity Oracle’s viable. And the simple truth is that platform players like Google or Force.com that can leverage an identity-rich business service that they also have are ideally suited to be trusted Identity Providers. But while a big platform player can certainly be a good Identity Provider, not all hosters should need to be Identity Providers to be successful. Instead, standards based identity services would be a great asset for hosters that want to be sticky (by being the best platform to deploy on) without having to take on the onerous task of being an Identity Provider (which has its own challenges) or passing on those responsibilities to their customers (which is what mostly happens today). And it would be an asset for SaaS vendors that want to have the freedom of choice that we all crave, and that want to be able to work with their customers identity infrastructure. As Dale says in the article:
You see, people can move an application from one host to another without much trouble.
Now, isn’t that a good thing, and something that we should be aiming for?
I believe that being a trusted IdP has nothing to do with hosters or SaaS. If others (and consumers in particular) are to trust an IdP, factors other than PaaS or SaaS come into play – like security of their data/infrastructure, security of their authentication mechanism, etc. A real IdP is almost like the root of a EV-cert – it is an onerous job to vouch for the identity of a person over net. For these reasons, I believe that we are still at a stage where the SaaS installation will have to work with enterprise IdPs – not the other way.
Great distinction re: Identity Hub vs. Identity Store. Top of mind for me as I wrote that were consumer sites like Mint.com, but clearly the PaaS “hosters” would need to be able to accept claims generated within the enterprise and similar services in order to be competitive in the future state.