metadata-driven attribute definition

[Cantor, Scott]

On 7/15/19, 10:30 AM, "users on behalf of Joshua Dachman" <users-bounces at shibboleth.net on behalf of jdachman at gmail.com> wrote:

> It seems to me that I asked a "how" question and the response I'm getting is "why do that?"

You got both.
 
> I've explained the why--to have a purely metadata-driven approach. To give some more context to that, I'm dealing with
> hundreds of SP's here (and the  number is constantly growing), each with different requirements regarding the names of
> the attributes they expect.

I have hundreds of SPs, and I have a handful of exceptions. The reason this doesn't come up as a request very often is that it's meant to be, and in practice should be, quite rare.

> I was hoping a purely metadata-driven approach could be wired up, but the impression I'm getting is that it will require
> development work. If you have any guidance on how to do that, please let me know.

There's no obvious way to do it in V3, the interface to the attribute encoding interface is very limited and has access to very little of the context of a request.
 
The new transcoding service in V4 is more amenable to allowing this, but it hasn't been implemented to this point to support that specific feature. It probably wouldn't be a lot of work to add it if a RFE is filed, but the actual configuration probably wouldn't be that improved since you'd have to configure it with an EntityAttribute tag name to look at for the name, instead of just outright configuring the name itself to use as a custom mapping. Some might prefer that I suppose.

-- Scott