Re: [kitten] Use of GSS_Get_name_attribute() to obtain further attributes

[Benjamin Kaduk <[e-mail]>]

On Mon, 13 Apr 2015, Simo Sorce wrote:

> On Mon, 2015-04-13 at 10:25 +0200, Alejandro Perez Mendez wrote:
>
> The advantage of a new API is that applications can use
> GSS_Get_name_attribute() with the implicit knowledge that the call will
> not block or generate traffic, however I guess that apps can list all
> known names ahead and then call the function only if the name is listed.
> So perhaps 1) is fine.

Let's see what sort of context we can put things in.

RFC 2743 explicitly calls out that the context-establishment and -deletion
calls may block pending network interactions, and the credential
management calls are also permitted to block on the network.  In contrast,
per-message calls are explicitly declared to not block pending network
interactions.

The word "block" does not appear anywhere in RFC 6680.  It would be nice
to hear from some of the original authors of 6680 about whether this was a
question they had considered.

I do not think I am opposed to (1) (i.e., letting GSS_Get_name_attribute()
block on network interaction), but if we proceed down that route, I think
we should file an erratum against 6880 to that effect.

-Ben