Re: service names - one registry vs. two registries

[Nicolas Williams <Nicolas.Williams at sun.com>]

I think a single registry is better than N similar ones, even though one
could always impose rules on the alternative such that collisions are
minimized.

I would also add that there are other service name registries in the
IANA that you should consider folding in here.  The one I'm thinking of
is the GSS-API service name registry.

Ideally what I'd like to see is something like this:


  +------------+--------+-------+-------+-------+
  |Service     | SRV    |Default|Aliases|is GSS?|
  | Name       |label   | Port  |       |       |
  +------------+--------+-------+-------+-------+
  | foo        | _foo   | 12345 |<none> | No    |
  +------------+--------+-------+-------+-------+
  | bar        |        | 23456 |foobar | Yes   |
  +------------+--------+-------+-------+-------+
  | ...        | ...    | ...   |...    | ...   |
  +------------+--------+-------+-------+-------+


Perhaps a SQL model would help: ?

CREATE TABLE services (canon_name TEXT PRIMARY KEY UNIQUE,
		       srv_label  TEXT UNIQUE,
		       def_port   INTEGER UNIQUE CHECK(def_port < 32768),
		       is_GSS     BOOLEAN
			   CHECK(is_gss = 'true' OR is_gss = 'false'));

CREATE TABLE aliases  (alias_name TEXT PRIMARY KEY UNIQUE,
		       canon_name TEXT,
		       FOREIGN KEY (canon_name)
		       REFERENCES services (canon_name));

(I didn't include CHECKs for name and srv_label, but, you get the
point.  I'm using SQLite3 syntax here, specificially w.r.t. type names.
I also didn't capture the need for aliases to be unique relative to
service canonical names.)

Nico
--