Re: Updating IANA "Operating System Names" registry

[John C Klensin <[e-mail]>]

--On Tuesday, January 26, 2010 13:15 -0500 Thomas Narten
<[e-mail]> wrote:

>> The IANA registry "Operating System Names" found at
>> http://www.iana.org/assignments/operating-system-names is
>> missing a few recent common OSes.
> 
> More to the point, who uses this registry and who cares?  It
> is old, and probably stopped serving a useful purpose years
> ago.
> 
> What protocols make use of the registry?

Sigh.

The FTP SYST command, among other things.   SYST is much less
important now than it was a few decades ago when there was a lot
more diversity of hardware architecture and operating systems,
but there is what we would now describe as a normative reference
from RFC 959 (STD 9) to this registry.

The DNS HINFO record is also supposed to use them.  While I
haven't see a lot of uses of HINFO recently in the public DNS,
we haven't deprecated the RR type either.

I have generated an errata entry for RFC 959 noting that SYST
should point to this registry rather than pointing more vaguely
to "Assigned Numbers".  I will leave similar comments on the
definition of HINFO and possibly a modification to the new FTP
Commands and Extensions registry (to note that this registry
specifies the parameter space for SYST) to someone who has more
energy and time for that kind of work.

With a quarter-century's hindsight, it is clear that this
registry, and the names in it, should have used structured names
of some style such as 

   OperatingSystem Delimiter Version Delimiter Subversion

possibly with explicit registration for only the first element,
but it is a little late now.   Getting the list up to date and
into shape would clearly require significant work, but I don't
think we should discourage new registrations by anyone who
thinks that particular entries are needed.

> Absent a real usage, the registry should either be shutdown,
> or just quietly left alone.

I don't see how one can shut down a registry that is used
normatively by two full standards (FTP and DNS) without first
deprecating the FTP Command and the DNS RR that use it.
Continuing the long-standing practice of neglect unless new
entries are actually needed seems entirely reasonable and has
the advantage of not requiring any action.  In general, it is
probably better to believe that registries should be updated
when someone has a substantive need for new entries than to try
to create some abstract requirement to add entries for their own
sake.

I am tempted to ask why we are even wasting time discussing this
registry, but I won't.

      john