IdP 3.4: Deprecated items in custom Syntax

[Cantor, Scott]

On 9/26/17, 6:08 AM, "dev on behalf of Takeshi NISHIMURA" <dev-bounces at shibboleth.net on behalf of takeshi at nii.ac.jp> wrote:

> We have experienced troubles about old format and new format of attribute-resolver.xml. For example, our example snippet in
> old format cannot apply to attribute-resolver.xml in 3.3.1.

As Rod said, that's a bug.
 
> So we need a conversion guide from old format to new format. Newer
> deployers can use this guide to convert snippets before applying them to
> their system.

The guide for 3.4 is "fix warnings", and the warnings should generally make it pretty clear what the replacement is. It's much easier to document that in notes on a page referencing the particular thing that's been changed than to try and combine it all into one document that won't be as correct or likely to be maintained well.

> As for deprecated features, we need to clarify "non-deprecated feature" in the following text:
> I know ad:SAML2NameID is deprecated but also exists in resolver: namespace.

The statement doesn't say anything that's deprecated won't be in that namespace, but that anything that isn't deprecated is in that namespace. Those aren't the same thing.

> There may be a timing issue (it was not deprecated til 3.3 release) or
> there may be two types of deprecated features.

There are not.

> Would all currently deprecated features (and possibly in future but before 4.0) be removed from 4.0?

There is no guarantee of that. We don't make every decision years ahead of time, it has to be taken on at the time so we're basing the decision on up to date information about usage.

A good example is whether we actually end up removing the resolver support for computed IDs generated inside attribute values. Would I use that feature? Not a chance. But that doesn't mean we absolutely know we're removing it.

> Another method come to my mind is labeling "DEPRECATED3.4" or so in each deprecated feature page. A deprecated list can be
> generated using these labels.

I believe pages can have labels, but I don't know if specific content in a page can.

-- Scott