Shibboleth Developers

[Leif Johansson <>]

On 02/02/2010 11:04 PM, Scott Cantor wrote:
> Etan Weintraub wrote on 2010-02-02:
> > I understand...and I'm not sure is my answer. I can see it making it easier
> > for initial deployments, but I think when I have to add attributes later, I
> > see myself forgetting that I have to modify attribute-encoder.xml also,
> > hence why I wanted the option of keeping it in the single file.
>
> I tend to believe it's the sort of thing that helps people initially and then 
> becomes a cause of errors later for the reason you're suggesting.
>
> I would say if it's simple and maintainable (in the code) to support both, 
> that's fine, but I always err on the side of the ongoing usage needs.
>
> An analogy in the SP is that it was tricky to come up with a filtering 
> stategy that reused the IdP filter syntax but didn't require that adding 
> attributes to the mapping file also require adding them to the filter policy, 
> but I felt very strongly that it would be a disaster to do that.
>
> My other suggestion to deal with clutter is to use defaults and bump some 
> things up onto the attribute definitions to enable instantiation of decoders 
> for the common cases, so that only some of them actually need explicit 
> declarations.
>
> -- Scott


Why can't attribute-encoders.xml be a single template applied by regexp 
to all attributes?

        Cheers Leif