Re: [port-srv-reg] - Getting DCCP port/SC rules right

[Gorry Fairhurst <[e-mail]>]

Thanks Joe - we still have one DCCP "invertor" in the rules, see below.

Gorry

Joe Touch wrote:
> I just posted a version including Magnus' comments. Attached is a
> version that further includes Gorry's below. See notes...
> 
> Joe
> 
> Gorry Fairhurst wrote:
>> Looks good to me,
>>
>> A few comments on the slides and the draft...
>>
>> Slide 1:
>> - It would be good to include the draft name on the slide:
>> draft-ietf-tsvwg-iana-ports-03
> 
> Done.
> 
>> Slide 8:
>> This line took me a wee moment to decipher:
>> "Port only for transports requested"
>> - is it better as?:
>> "Port only for requested transports"
> 
> Done.
> 
>> Slide 11:
>> This line doesn't seem to be the whole story:
>>
>> "Service code to at most one port"
>> -> This isn't a requirement as per RFC 5595 (e.g. section 4.1), although
>> it is good as common practice, unless the requester knows better and
>> requires expert review (as described in the draft section 10.3.2). Did
>> you mean at most one port per service code?
> 
> Yes. (sorry - I wrote it backwards, using "to").
> Fixed.
> 
>> =============
>>
>> And a wee bit of (late) feedback on the DCCP topics in the draft:
> 
> I'll note this for the update, but some other notes as well:
> 
>> Section 10.3.1:
>> - It is good to also cite RFC 5595 (you currently cite the draft) for
>> service code information, since this updated RFC 4340 in many ways.
> 
> Will do.
> 
>> - We MUST note in the new draft that no registry allocations can be made
>> for zero or 4294967295, and no registry allocations can be made for
>> 1056964608-1073741823 (high byte ASCII "?") reserved for private use
>> [19.8 of RFC 4340].
> 
> The doc just talks about coordinating tables. It doesn't repeat rules
> that occur elsewhere (or at least shouldn't, and if it does they should
> be removed).
> 
>> - I think we may have missed the rule that:
>>   "IANA MUST NOT allocate more than one DCCP server port with a
>>    single Service Code value."
> 
> see 10.3.2:
> 
>    o  IANA MUST NOT allocate a single Service Code value to more than
>       one DCCP server port.

That seems the wrong way round in the draft you are editing.

The fundamental rule is: ONLY port one allocated to a particular SC.

The recommended registration policy needs to try to allocate an unused 
port (for DCCP) to a new SC, avoiding unintentional allocation of one 
port to more than one SC. (Note thought that this *IS* permitted in 
DCCP. When you need multiple SC per port, an allocation CAN be requested 
with expert review).

> Is this the same thing or not? If so, which wording is better? If not,
> do we need both, and how can we make this more clear?
> 
>> - Does this draft also updates RFC 5595, since the latter states where
>> ports and service codes are registered?
> 
> Will do.
> 
>> Section 10.3.2:
>> - I think it would also be helpful to cite RFC 5595 in 10.3.2 - since
>> this explains how ports and SC interact and how you can manage without
>> port allocations.
> 
> Will do.
> 
>> Section 11: You didn't fix the acknowledgments:-)
>> - I don't recall Tom being in the loop, what I recall was that I ripped
>> the first version of text from an earlier draft of RFC 5595
>> (draft-ietf-dccp-serv-codes-04.txt).  You could check with Tom, if you
>> like;-)
> 
> Correct - we didn't fix the acks. Let's do that in the next round ;-)
> 
> Joe
> 
>> Gorry
>>
>>
>> Joe Touch wrote:
>>> Feedback appreciated. I will be able to edit until about 0800 Monday,
>>> but need to post for those online by then.
>>>
>>> Joe
>>>
>>> Stuart Cheshire wrote:
>>>> On 29 Oct 2009, at 7:27, Joe Touch wrote:
>>>>
>>>>> I'm open Mon AM and can present if useful. I'll also be at TSVWG.
>>>>>
>>>>> Joe
>>>> Do you have some slides prepared Joe?
>>>>
>>>> If you do, and you can email them to the list, I'll take a look at them
>>>> tonight and give you feedback.
>>>>
>>>> Stuart Cheshire <[e-mail]>
>>>> * Wizard Without Portfolio, Apple Inc.
>>>> * Internet Architecture Board
>>>> * www.stuartcheshire.org
>>>>
>>>> ------------------------------------------------------------------------
>>>>
>>>> _______________________________________________
>>>> Port-srv-reg mailing list
>>>> [e-mail]
>>>> https://www.ietf.org/mailman/listinfo/port-srv-reg