Monday, March 24, 2008
It's unsanitary, Kim!In a blog entry today, Kim Cameron both puts words in my mouth and twists the ones that come out to serve his "straw man" purpose.
In commenting on my recent post about the death of the metadirectory, he says: "Who would want to get in the way of Dave’s metaphors? He’s on a streak. But he’s making a fundamental mistake, taking an extreme position that is uncharacteristically naive."
What did I do? I advocated the virtual directory as the better vehicle for all of the ID data needed in the SaaS world.
Kim implies that, somehow, I called for the virtual directory to be authoritative. That's simply not so. the virtual directory is merely the conduit to the authoritative source, wherever it might be. The application developer doesn't even need to know the authoritative source of the data - or need to re-write code if that source changes.
But then he goes on to say: "Application developers like to use databases and tables. They have become expert at doing joins across tables and objects to produce quite magical results. As people and things become truly first class objects in our applications, developers will want even more to include them in their databases."
I couldn't agree more. As a developer, I always prefer to have a local cache of the data I need in a (for me) easily manipulated data structure. But that does not mitigate against the use of a virtual directory. Far from it. The application database (for those who cling to it like Linus and his blanket) now can serve two purposes - one to subscribe to virtual directory data and one to publish!
The application database is the authoritative source of the application-generated data, and should be linked to the virtual directory which will consume this data and make it available for other applications and services. At the same time, any data which the application consumes - but which it is not authoritative for - can be populated at run-time from the virtual directory. For the developer who thinks this is a performance hit (and for whom accuracy is less important than an extra millisecond), a "synchronization stored procedure" would handle data changes without stealing precious time from the user-application interaction. It really is win-win.
Now the argument could be made that a synchronization engine (such as in a provisioning system) could periodically update all of the various datastores with any new or changed identity data, but that simply takes the well-known synchronization problems of the metadirectory and magnifies them by the dozens, hundreds or thousands of application datastores within the organization. That's a recipe for disaster. If an individual developer, for an individual application, wishes to sacrifice accuracy and risk the potential of error caused by out-dated data, or data whose location has changed in the hope of a spurious speed improvement (almost immediately unnoticeable due to the fluctuating nature of network thruput), they'll quickly learn, I believe, that "haste makes waste."
The further error Kim makes, though, is to believe that a virtual directory can't look like a SQL database to the application (or an XML database for web services developers). The folks at Radiant Logic would certainly disagree. It's all about the context. I'd invite Kim, and other skeptics, to our sessions on Identity and Context (including one about context and user-centric identity, as well as context and virtual directories) at next month's European Identity Conference in Munich.
Comments: Post a Comment
© 2003-2006 The Virtual Quill, All Rights Reserved Home