Re: [kitten] RFC2743 errata 4251

[Nico Williams <[e-mail]>]

On Sat, Nov 08, 2014 at 02:48:20AM +0100, Martin Rex wrote:
> Nico Williams wrote:
> > I wrote this yesterday but neglected to click the submit button.  Just
> > submitted:
> > 
> > http://www.rfc-editor.org/errata_search.php?eid=4151
> 
> I hate to do this (to you), but I'm "violently" opposed to this errata.

Hey, at least you hated doing it!

> I believe you have misunderstood the meaning of the original text.

Perhaps.  The goal is to explain how to process an unexpected error
token.

> Quoting from your errata:
> 
>   Section 2.2.4 says:
> 
>    o  GSS_S_FAILURE indicates that the context is recognized, but that
>    the GSS_Process_context_token() operation could not be performed for
>    reasons unspecified at the GSS-API level.
> 
>   It should say:
> 
>    o  GSS_S_FAILURE indicates that the peer had an error consuming the
>    last context token sent to it (when the local side must have been
>    fully established but the peer hadn't yet been).  The minor status
>    code provides error information from the input token.
> 
> You find similar descriptions for GSS_S_FAILURE semantics all over
> the GSS-API specification _on_purpose_, and you _may_not_ remove that
> meaning here.

Fair enough, we'll add text rather than replace text.

> You may at most add a hint that the minor_status(!!) that must
> accompany a GSS_S_FAILURE major_status returned from
> gss_process_context_token() may not just cover reasons that are
> a "local matter" (i.e. implementation issue of the current gssapi
> mechanism) but include reasons conveyed through an error token
> from the remote side.  It will be impossible to define portable
> behaviour to programmatically react to specific such reasons,
> because what will be returned is not just a local issue, but
> a remote plus local issue combined.

I agree.

> Btw. any minor_status that conveys an error from a remote peer that
> was carried in an error token ought to clearly say that this
> information came from the remote peer (and is not a local error) in
> the textual translation from gss_display_status().

Indeed.

Nico
--