Liberty Alliance Announces openLiberty Project

A few comments after having spent a few months working on Liberty stuff last year.

The Liberty Alliance identity stuff basically allows web sites (and why not other systems as well) to share "common user databases". So each and every web site does not have to make users register and remember passwords. They can refer the user to some identity provider for authentication (the technology does not force one to use any specific identity provider). The protocols allow pretty much anything from purely anonymous authentication to sharing user attributes with the users permission (and much much more if all the protocols are implemented).

All in all, it is a really nice concept, but from the perspective of free software people, it is very business oriented. I think that stuff is good and useful for adoption, but one needs to be able to get through all that business jargon.

As for this specific implementation. This is not something new. There already is a really decent free implementation about the Liberty Alliance basic specs: Lasso by Entr'ouvert. I can really recommend this package if people end up experimenting with Liberty. (Disclaimer: I am not affiliated with Entr'ouvert.)

I have had bad experiences with hyping the words "open source" with Liberty stuff. For example SourceID is open source and it is kind of a reference implementation for the ID-FF spec, but it is almost entirely undocumented and poorly debugged (a colleague found out it worked with only one user account, imagine if a Kerberos implementation worked only with one user id), and it is almost useless for any real use other than demos. (Incidentally SourceID was included recently on the list of "enterprise" software as seen on LWN :-)