Re: [port-srv-reg] Four final points for discussion

[Joe Touch <[e-mail]>]

Hi, Stuart,

On 9/20/2010 9:08 PM, Stuart Cheshire wrote:
> On 16 Sep, 2010, at 09:44, Joe Touch wrote:
>
>> Any time more than one string maps to the same port number it's a kind
>> of an 'alias'. The only way for the client to know the difference is
>> via in-band information.
>>
>> I.e., we would allow aliases for:
>> - different services
>> - that can be resolved in-band
>>
>> Unless BOTH of those apply, we would not allow aliases except:
>>
>> a) legacy (www, http)
>> b) to support changeover to the new namespace
>>
>> Joe
>
>
> I think we agree on the concept, but the terminology previously confused
> me.
>
> When you said, "alias," I read that as meaning, "different names for the
> same service."
>
> Now I see that when you say, "alias," you mean, "different services on
> the same port." (I'd call that a "collision", except in the special case
> of a new service which is a compatible extension to an old one on the
> same port.)

Please see the text Lars just checked-in.

- overloading is when different service names correspond to the same 
port number, of which there are three cases:
	1- extended services with in-band resolution mechanisms
	2- legacy names for the same service
	3- names this doc phases out for syntax reasons (i.e.,
	instances of #2 this doc deliberately creates)

The current text does not use a single word to describe 1 vs 2 or 3; it 
distinguishes them in prose instead.

(The only further mod might be to remove the "Overloading occurs" in the 
first bullet, which might suggest (inadvertently) that overloading 
doesn't occur in the other two bullets; it's really in all three, as per 
the lead-in intro to the bullets).

Joe