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

[Joe Touch <[e-mail]>]

Stuart Cheshire wrote:
> I have four final points for discussion
> 
> 1. Name of the Registry
> 
> Right now the document calls it:
> 
>    Transport Protocol Port Number and Service Name Registry
> 
> However, Section 5 says:
> 
>    Service names are the unique key in the Port and Service Name registry.
> 
> Since Service names really are the primary identifier, and the port number is optional, and we expect to see more and more registrations without port numbers, would it make sense to switch the order of the words, and make it:
> 
>    Registry of Service Names and Transport Protocol Port Numbers

Sounds better.

> 
> 2. Service Name Rules
> 
> I liked Joe's earlier suggestion to disallow all-numeric service
> names, to avoid service names that look like a numeric port number.
> However, even with that rule, we still allow service names like this:
> "6000-6063" (looks like the X Window System port range). Do we care? We
> could prevent that by requiring that all service names contain at least
> one alphabetic character.

The point is only that a service name cannot be confused with a port
number and used in its place without lookup.

If we want to require an alpha - which would be fine - both the English
and BNF would need to be revised, AND we would need to re-scrub the
current table and create a new list of corrective aliases for legacy
names that wouldn't work, if any (I don't think there are).

> 3. Inconsistent terminology.
> 
> We use the term "Registered Ports" in some places to mean "ports in
> the range 1024-49151", and in other places to mean any "port recorded by
> IANA in the Registry".

The problem arises from numerous previous documents that refer to the
ports as quoted:

> For example, the first meaning:
> 
>           <t>the Well Known Ports, also known as the System Ports, from 0-1023
>           (assigned by IANA)</t>
> 
>           <t>the Registered Ports, also known as the User Ports, from
>           1024-49151 (assigned by IANA)</t>
> 
>           <t>the Dynamic Ports, also known as the Private Ports, from
>           49152-65535 (never assigned)</t>

We can't fix the problem with the word "registered". We ought to avoid
using it for the IANA process.

Perhaps the word "reserved" or "assigned" for the IANA process?

I.e.:	IANA Reservations / Assignments
	reserved ports (0-49151)

> and now the second:
> 
>       <t>It is important to note that ownership of registered port numbers and
>       service names remains with IANA. For protocols developed by IETF working
> 
>       Thousands of applications and application-level
>       protocols have registered ports and service names for their use, and
>       there is every reason to believe that this trend will continue into the
>       future.
> 
> Would it be better to strictly use the terms "System Ports" and "User
> Ports" to denote the ranges,

Yes. That would help reduce the ambiguity.

> and keep the term "registered port" to just
> mean generically, "recorded by IANA in the Registry"?

We can't use that term; it's already out there to mean 'user ports'. We
should use a different term - either as suggested above or something else.

> 4. Aliases
> 
> The document says:
> 
>       <t>IANA is also instructed to indicate which service name aliases in the
>       existing registry are the primary aliases (see <xref target="srvname"/>).</t>
> 
> Why should we burden IANA with this decision making? I bet there
> aren't that many. Let's just work it out ourselves, and list them in the
> document. I'll grep the IANA ports page and send a followup email with
> the list of aliased names.

Agreed.

Joe