Re: [port-srv-reg] Updating the draft with WGLC comments

[Magnus Westerlund <[e-mail]>]

Joe Touch skrev 2010-11-18 01:45:
> Hi, Magnus,
> 
> The feedback from Paul suggests it would be useful to update Sec 7.
> 
> Despite the explicit warning - already in the doc - that these 
> principles are NOT binding, it might be useful to discuss the issue of 
> whether separate ports should be allocated for requests for new protocols.
> 
> I.e., http vs https is currently legacy. We already expect that new 
> requests for nonsecure legacy services could result in a new, secure port.
> 
> The question is whether a brand new service should be allocated separate 
> ports for secure and nonsecure variants.
> 
> The document discusses this point as follows:
> 
>     o  IANA will allocate only one assigned port number for all versions
>        of a service (e.g., running the service with or without a security
>        mechanism, or for updated variants of a service)
> 
> ...
>    - Further,
>     previous separation of protocol variants based on security
>     capabilities (e.g., HTTP on TCP port 80 vs. HTTPS on TCP port 443) is
>     not recommended for new protocols, because all new protocols should
>     be security-capable and capable of negotiating the use of security
>     in-band.
> 
> Here's the TLS summary
> 	for:
> 		Mike D'Errico
> 		Nico Williams
> 	against:
> 		Paul Hoffman
> 		Marsh Ray - really just wants default to secure
> 		Richard Hartman
> 
> Some just wanted security all the time:
> 	Geoffry Keating
> 	Mike D'Errico
> 
> I didn't see that they came to consensus on this issue. We can easily 
> omit the security text altogether from this text if preferred, and let 
> the TLS community make a final BCP recommendation.
> 
> However, despite their status as security experts, I find their logic 
> disturbing. Port numbers themselves have no inherent security, so 
> ultimately only the application can require a service to be secure 
> anyway. Using port number blocking to assume security is laughable at 
> best, so I stand by the current text.
> 
> IMO we already have enough wiggle words that this section isn't binding 
> anyway. IMO, let the TLS folk create a BCP to the contrary, at which 
> point some of us (me at least) will write a doc explaining why port 
> numbers aren't security anyway ;-)

My view is that there seem to be no real security benefit from running
separate ports generally. One anyway has to live with the downgrade
attacks etc. Thus I think port space preservation is still the main goal.

> 
> Thoughts? Leave it? Take it out because a non-consensus subset disagrees?
> 

I would do some minor tweaks, at least to the following sentence:

IANA will allocate only one assigned port number for all versions
of a service (e.g., running the service with or without a security
mechanism, or for updated variants of a service)

People interpret the "will allocate only" very strict. I think we can
reword this to be one degree less strict. like:

IANA will with extremely few exceptions allocate only one assigned port
number for all versions of a service (e.g., running the service with or
without a security mechanism, or for updated variants of a service)


Cheers

Magnus Westerlund

----------------------------------------------------------------------
Multimedia Technologies, Ericsson Research EAB/TVM
----------------------------------------------------------------------
Ericsson AB                | Phone  [phone]
Färögatan 6                | Mobile [phone]
SE-164 80 Stockholm, Sweden| mailto: [e-mail]
----------------------------------------------------------------------