Re: [port-srv-reg] we need to make progress

[Joe Touch <[e-mail]>]

On 9/4/2010 12:33 AM, Stuart Cheshire wrote:
> On 3 Sep 2010, at 10:07, Joe Touch wrote:
>
>>> This implies that a given service name can have *different* port
>>> numbers assigned for different transport protocols.
>>
>> That is correct, and always has been.
>
> Can you give an example?

DNS and NFS. We review requests regularly for other numbers that do the 
same (except for formatting differences to deal with transport 
differences) in port reviews for IANA.
...
> The only example I know that people tend to cite is DNS, and even
> that is not strictly the same application protocol over different
> transport protocols -- over TCP DNS has a two-byte length; over UDP
> it does not.

The "same service" doesn't require the same packet format. Sequence
numbers and other information needed to deal with UDP's unreliable 
nature (lossy, out of order, can duplicate) are often omitted, and some 
framing information may be added to deal with TCP's lack of framing.

<> I don't like the idea of a primary alias either; there's no notion of it
>> in the way users interact with port indices (/etc/ports or SRVs).
>
> Yes, there is. Safari browses for _http._tcp. Nothing browses for _www._tcp.

We only know that nobody registered "www" in your database; we cannot 
know whether there are no apps that browse for this string.

>> Agreed. However, note that the document itself creates a number of
>> aliases for names that don't follow the new syntax.
>
> And the new names are all the "primary" name to be used in SRV records and similar.

IMO, there is no such thing as primary. It's ALL and ANY - register ALL, 
lookup ANY. That's the only way we know things will work.

If we are to continue to use the term "primary", we need to define what 
it is and what it means; IMO, that's a can of worms we should avoid.

Joe