Reverse of IdP Discovery Service

[Peter Schober]

* Rod Widdowson <rdw at steadingsoftware.com> [2018-01-26 10:12]:
> > At least that's how I create pages like this,
> > https://eduid.at/entities/sp (German UI but should be self-evident).
> 
> That is a really cool page !  Thanks Peter

Thanks, but -- as so often is the case for me -- all credit needs to
go to SWAMID from whose (old) wiki I took the inspiration (and
possibly even some of its initial implementation, as you could see the
XSLT used from the old NDN/SWAMID wiki) years ao.

I have dabbled with including the logos, too
https://eduid.at/entities/idp/logos
https://eduid.at/entities/sp/logos
but screen estate is already limited as it is, so I don't do that atm.

Note that the version at https://wayf.aco.net/smev/ allows selection
of a few other metadata aggregates (the list of which is not
religiously kept up tp date) and prefers xml:lang="en" over
xml:lang="de" (and uses english strings in the HTML UI, too), so may
be somewhat useful to others.
While both look for data in all the canonical places and in the "Best
Practices" order (mdui:DisplayName over md:ServiceName, likewise for
descriptions, etc.) and both have sortable columns the SMEV version
above currently also has a few more features:

* shows selected Entity Categories (R&S, CoCo, Hide[fromDisco])

* has a javascript "search" field in the upper right to limit the
  shown entities to any matching strings. E.g. enter "R&S" without the
  quotes into the search fieled to limit SPs to R&S (or IDPs to IDPs
  supporting R&S, if you're on the IDP page).

The dynamic (in-browser) search is more useful than immediately
obvious, as the HTML also has more info than is always visible:
E.g. you can mouse-over an attribute's friendlyName (comes from
metadata, so may not always be shown consistently) to get the formal
attribute name (again, everything is from metadata).
That in turn means you could limit the entities shown to those that
mention a specific (part of an) URI in their requested attributes, by
entering e.g. "1.1.6" into the search field, as shortcut for ePPN. And
then maybe sort the "limited" result after some column, say, Entity
Category (click column head twice to sort non-empty first), that kind
of thing.

Of course since there's no local data and no caching (except keeping
fetched metadata around for a short time, to avoid DoS'ing the
metadata URLs) and all XSLT processing happen on-access time, and all
sorting in the browser, this all basically breaks down for larger
aggregates: Rending the HTML for the eduGAIN aggregate takes ~30
seconds on an old notebook here, when venture capital tells us 3
seconds rendering means our business will be gone (or whatever).

-peter