An open discussion list for topics related to the eduGAIN interfederation service.

[Peter Schober <peter.schober AT univie.ac.at>]

* Stefan Winter <stefan.winter AT restena.lu> [2019-09-27 08:02]:
> If this freeze-in-time (and thus not being able to use empty lang at
> all - which we just heard might actually make sense in the case of
> "Logo") is what we want [...]

We must be reading different things, then, in case you're referring to
Scott's reply. (To my eyes) Scott certainly did not say that having an
empty xml:lang="" on mdui:Logo elements might "make sense": In every
such case (that allows for absence of a language value) you'd simply
leave out/away the xml:lang instead.

The minor -- and quite peripheral to this discussion -- point about
the singular corner case of mdui:Logo (that has an /optional/ xml:lang
in its spec) alone was worth allowing for xml:lang="" within the
Shibboleth SP software implementation (which it already does, I just
tested this, so nothing to do/change here), *not* that empty
mdui:Logo/@xml:lang="" attribute values "made sense".

To summarise the relevant points about the errors I reported on
this list:

* Some SAML schemas *require* xml:lang to be *present* in some places,
  e.g., mdui:DisplayName.

* Having an empty string value for xml:lang="" does *not* satisfy that
  requirement (as that would mean "treat it as if it were absent" when
  the schema does require it to be there).

So you can stop right there and declare those cases as invalid.
(Whether a given piece of software catches those currently or not.)
And without talk about locking outselfs to certain XML schema versions
or obsolete software or anything else.

-peter