Re: [art] Revising BCP56: On the use of HTTP as a Substrate

["Roy T. Fielding" <[e-mail]>]

> On Jul 16, 2017, at 5:41 AM, Ira McDonald <[e-mail]> wrote:
> 
> Hi Martin,
> 
> I'll try to be more specific:
> 
> IPP uses a unique binary encoding of its primitive datatypes (integer,
> language-tagged string, keyword (registered) as distinct from name
> (vendor free-form), etc.).  So the HTTP payload of IPP is opaque.
> 
> IPP original authors were told that using an HTTP substrate would make 
> penetrating firewalls easier - it didn't of course.  

Umm, what advice are you referring to?  I remember

  ftp://ftp.pwg.org/pub/pwg/ipp/minutes/961212-ietf-bof.txt

and being quite explicit about why that plan was bogus. I also remember
advising that HTTP be used as designed, not tunneled through with an
opaque blob of poorly designed vendor-specific options.

Whatever problems IPP may have encountered after that, they weren't due
to the authors following IETF advice, nor were they due to using HTTP
as a substrate (any more than they were due to using TCP as a substrate).
They might have been due to not inventing a completely new protocol that
performed streaming RPC with opaque data through firewalls, but I think
that's a bit fanciful given the lack of successful examples.

....Roy