ACSUiInfo needs to filter against empty names as well as empty languages

Id forgotten about the schema fix (if all you have is a hammer .... so I looked for a nail).

Much more elegant

There are challenges trying to do this in the objects themselves because of intermediate states (it would have to be in the unmarshalling process, not the implementation objects).

When I did it in C++ it was more effective to use the Validator construct that used to exist in Java, and apply rules to the objects after the fact so the code could run them under the correct conditions.

Short of reimplementing all that again, I think the simplest option is just to alter the local schemas to impose min facets on the string and anyURI types we use selectively. Most people don't validate so it won't change much but those that do could get more strict behavior.

But we could rebuild the validation layer, it isn't that much code, just a lot of grunt work. I had it much easier because of macros.

I'm not going to reopen this (yes), But I'll observe that this fix was topical to the IdP.

I wonder whether the supplementary fix is in OpenSAML. 

There are obviously some SAML elements (and hence classes which implement org.opensaml.core.xml.schema.XSString) where empty strings are invalid and we could have better locality to warn?

Similarly {{ org.opensaml.core.xml.LangBearing}} should perhaps often (always) not have an empty language.

I'm wondering whether we could add new classes which are intolerant of emptyness for

* LocalizedURIImpl (for lang)

• LocalizedName (for lang)

* XSStringImpl (for the contents).

and derive the required impl classes from this?

Then we'd not need to have the null checking when we mine the SAML structures.

I don't speak SAML well enough to know which elements are the ones which matter, otherwise I'd make a concrete proposal,.

Tests extended

Fixed other outliers (found by grepping of xml:lang) being

• UiInfo

• Organization

I'm pretty sure that this is done, but I'll leae open pending nightlies and another run over the code.

It says AssertionConsumerService but I think we're talking about AttributeConsumingService?

Almost certainly

What about the case where both the lang and the content are empty?

We already filtered on empty language. Empty text was what was missing..

I'm guessing there are no unit tests in this area?

That would surprise me since I’m pretty sure I authored the code. Part of work to do before closing is to have a regression test..