Re: [kitten] draft-ietf-kitten-rfc5653bis-02 review

[Nico Williams <[e-mail]>]

On Sun, Feb 15, 2015 at 10:21:08PM -0600, Nico Williams wrote:
> That said, the use of streams strikes me as dangerous: in the presence
> of a self-framing mechanism, an application that passes in streams
> corresponding to the application protocol streams... would function
> correctly, whereas it would fail to work with non-self-framing
> mechanisms.
> 
> (The GSI mechanism that uses TLS is a self-framing mechanism.  When the
> application writes the tokens without further framing the result is TLS
> on the wire, even though a GSS interface is used!)

That is, I am in favor of deprecating the methods dealing with streams,
or at the very least adding text describing the framing issue and
recommending (lower-case?) the use of memory-based, one-token-only
streams.

On the other hand, I do want to add a mechanism attribute for indicating
that a mechanism's tokens are self-framed...

Nico
--