Re: [port-srv-reg] Comments on SVN revision 68

[Joe Touch <[e-mail]>]

On 10/4/2010 7:24 AM, Magnus Westerlund wrote:
...
>> ...
>>> 2. Related to section 5.1, what about the rule that there are no consecutive
>>> hypens?
>>> Could someone register a---------5 as a service name/port number?
>>> Are there any issues with that?
>>
>> They can now; under the new rules, they could not. I see no issues with
>> that.
>>
>> Sequential hyphens are confusing. So is the name "oooooooh" - you have
>> to count the 'o's and get it right.
>>
>> I don't see syntax as a place to specify how to avoid such confusion.
>> That can (and already does) happen via gentle suggestion by IANA reviewers.
>
> I agree that we don't need to exclude it in the syntax. A question is if
> we should include a recommendation against doubl hyphen and other cases
> of repetitions that aren't readable as regular english?

I see no need to do so.

FWIW, I had thought we already agreed to change the rules to prevent 
sequences of hyphens. Right now, the ABNF is correct in this regard, but 
the normative rules need another rule:

- hyphens are not allowed adjacent to other hyphens

>>> 12.  The following paragraph describes that we only assign the requested
>>> protocol and mark others 'Reserved':
>>>
>>> "The new allocation procedure conserves resources by allocating a port
>>> number to an application for only those transport protocols (TCP, UDP, SCTP
>>> and/or DCCP) it actually uses.  The port number will be marked as Reserved -
>>> instead of Assigned - in the port number registries of the other transport
>>> protocols."
>>>
>>> QUESTION: When reserving the non-used protocols, are we putting additional
>>> lines for each entry?  This will make it very clear however the registry
>>> will get much larger.  (and it is already quite large)
>>
>> There ought to be an explicit indication of reserved ports; this can be
>> accomplished several ways:
>>
>> 	1) separate lines
>> 	2) some other organization where one listing also indicates
>> 	which other protocols are reserved
>>
>> This is out of scope for the doc, though - the doc shouldn't focus on
>> how the info is represented in the registry.
>
> Yes, I think even a statement before the list saying that protocols for
> a given port number that has at least on registration are reserved but
                                         ^^^ = one
> not explicitly indicated will do the trick.

That works for humans, but not for systems using the table source as a 
database (screen-scraper or otherwise). For the latter, we need an entry 
for anything that is not "available".

Joe