Meta-directories? I'd say quaint, but not quite dead.

An interesting conversation is taking place within the blogsphere about meta-directories, with Dave Kearns and Kim Cameron on both sides of the argument. This was all inspired by a blog entry on the 4th of March from Jackson Shaw called "You won't have to kick me around anymore!". That musing was about HP's retreat from the identity management market, but makes a statement about meta-directory technology:

Let's be honest. The meta-directory is dead. Approaches that look like a meta-directory are dead. We talk about Identity 2.0 in the context of Web services and the evolution of digital identity but our infrastructure, enterprise identity "stuff" is decrepit and falling apart. I have visions of identity leprosy with this bit and that bit simply falling off because it was never built with Web services in mind.

I started in this area in 1993 and some of the same architectures are still out there.

First of all, to clarify what "meta-directory" means (at least, to me!). I am thinking about "Via" (Kim's baby, the product that Microsoft acquired in 1999 together with Kim's company, Zoomit). I'm also thinking about Novell Dir-XML, Siemens DirXmetahub and the Critical Path Meta-Directory Server. Old products, created many years ago. You don't really see much happening with this technology any more, because it has its share of problems, and unless assisted with other technologies, does not fit well into today's much more dynamic identity and access models. The only exception to that is probably MIIS, but I'll get to that in a minute.

The old traditional "meta-directory" technology works by creating one big "centralised directory" (or "metaverse" as it's known in MS-speak), pulling data from everywhere into that centralised directory and then pushing data out into all directions either. This approach is usually not a good fit by itself, because it has several significant shortcomings. I would not go as far as call the technology "dead" (it's impossible to ignore the many MIIS installations out there), but I'll call it something else: "quaint". Now that word has several meanings according to the dictionary, but I sure don't mean "marked by skillful design, beauty or elegance"!!!

Microsoft has made an investment into that technology by rewriting MIIS pretty much from scratch. And Siemens to this date probably has the most comprehensive and advanced meta-directory implementation with its DirXmetahub component that is part of its Dir-X offering. Nevertheless, meta-directories are arguably still around mostly because Microsoft forces this technology onto its customers for what I think are political reasons: Several people working for Microsoft in the field have told me that it was in Microsoft's interest to have Active Directory as a central component, and believe it against Microsoft's interest to have a "filtered access", such as a virtual directory in front of AD, abstracting information away from what should be the authoritative source. I never really understood this fear, but recently it seems that this brick wall may be slowly starting to crumble (see below).

Some experts in the field still obstinately (in my opinion) push meta-directory technology as the only way to integrate multiple sources of identity information. I think this is very short-sighted. This might have been true in the last century, which is not even that far ago. But in a truly dynamic environment, meta-directory technology and a "synchronisation-only" approach just tends to get into the way. Likewise, the idea that virtual directories by themselves could solve all integration issues is wrong. It's never been only one or the other, unless you had a specific problem to solve. It's not synchronisation or virtualisation. You need both, at least if you are in a dynamic identity environment, or have a vision to get there.

So what is the solution for the future? Some people believe that virtual directories will eventually fully supplant meta-directories. Coming from the virtual directory world myself (I worked for Symlabs before joining Kuppinger Cole), I never truly believed that - at least not the virtual directories that were around at that time. Virtual directories and meta-directories could co-exist, and the combination of both had in the past shown great benefits. Think of it as the screwdriver vs. the hammer. Sure, with some brute force you might argue that you can use a hammer to put a screw in, and with some agility you might use a screwdriver to hammer in a nail. But you're likely to damage something in the way, or at best, not be very practical about it.

I think the future is definitely in the convergence of traditional directory servers, virtual directories and synchronisation solutions to provide rock-solid dynamic directory infrastructure. To a certain extent we can already see this. Maxware (before getting acquired by SAP) and Radiant Logic have already released early, basic versions of synchronisation solutions that harness the power of virtualisation and combine synchronisation with dynamic, abstracted multiple views of data, rather than the static meta-directory approach.

In the future I believe we will see "super-directories" that combine traditional data storage with LDAP access, virtual views and synchronisation features. Some of the players in this space are gearing up to do this already. As synchronisation is usually well-established technology by most of the large players in the identity management space, the missing part is currently still virtualisation, and especially the integration of virtualisation and synchronisation.

Sun and the OpenLDAP foundation, for example, have already added some basic virtualisation features to their directory servers. Oracle has acquired OctetString a while back, and has arguably the most complete, all-around implementation of directory services, synchronisation and virtualisation. Novell, IBM and Microsoft are still lagging behind in this space, with some of the "old guard" defiantly resisting directory virtualisation and hanging on to last century's belief that synchronisation can solve everything. But there are signs that this resistance is crumbling. It better be. Recently, at DEC2008, Microsoft's Stuart Kwan presented Microsoft's vision of a truly dynamic identity infrastructure based on an "identity bus", where applications could plug in, and "transformers allow us to fold, spindle and mutilate the data in any way we want" - changing internal claims into any other format required by applications. Surely virtualisation is not the only piece that is needed to fulfill such vision, but it is an important (and still missing!) piece. Kim Cameron has not been known to be a big fan of virtual directories - and he still shows some scepticism for the "virtual only" approach, but seems to be warming to virtualisation in combination with synchronisation in one of his recent postings:

So we are led to the conclusion that we need a spectrum of synchronization and remote access capabilities. We should be able to use policy to define what information is stored where, and how to get to information that is not stored locally - e.g., combine metadirectory and virtual directory functionality.