RE: draft-ietf-impp-im-03.txt

["Peterson, Jon" <jon.peterson@neustar.biz>]

> [ color me no-ob, but i have comments.  iesg cc:d because the
>   author, the other aps ad, and some concerned parties are here ]
> 
> 3.4.1 The Message Operation
> 
> why are the semantically different failures all given the same
> response syntax "faulure?"  it would seem useful to let the
> sender know the reason for the failure, as they are clearly
> differentiated.
> 

Just speaking from my own perception on this, having some sort of 'failure
code' that differentiates various failure causes went beyond the 'minimal
interoperability' scope of the gateway model. No question cause codes would
be useful, and IM/presence protocols generally support them. But in order to
be compliant with the CPIM model, I don't know if a protocol should be
compelled support some way of expressing the reason for a failure.

As an aside, Dave Crocker argued at some length that success and failure
operations (which he compared to mandatory MDNs) create end-to-end chains of
dependency that undermine system reliability, and he therefore opposed them
categorically. So just from a WG politics angle, adding further complexity
to these operations was a little sticky.

I'll fix the nits below. Thanks for the good comments.

- J

> ---
> 
> nits:
> 
>    CPIM model.  When a client sends a operation to an instant 
> messaging
>                                   ^
> -
> 
>    messaging protocol must have some message or capability 
> that provides
>    the function described by all given operations)
> 
> s/all/each of the/
> 
> -
> 
> randy
> 
>