Re: iSCSI Naming: iqn format specification

[Mark Bakke <mbakke@cisco.com>]

Glen Turner wrote:
> 
> Mark Bakke wrote:
> >
> >    - Not everyone has or needs one for other purposes; might cause
> >      extra load on IANA-assigned name space, especially if end users,
> >      researchers, or university projects start applying for them.
> >      In the past year, IANA has assigned about 3,000 enterprise
> >      numbers.
> 
> I'd suggest altering the syntax to allow for OIDs, not just
> IANA-assigned enterprise numbers.  Otherwise people with ISO-
> assigned numbers are going to end up holding multiple OID
> allocations, which begins to be administratively nightmareish.
> 
> This also fixes the "load on IANA" problem.

Actually, the load on IANA problem is probably not all that
bad.  Most companies, particularly hardware or software manufacturers,
that want to protect themselves in this way already have an enterprise
number to use in their MIB.

> As a real example, OIDs are required for LDAP schema and each
> organisation can be expected to have a unique LDAP scehema (such
> as Example Corp having an examplePerson schema).  To reduce the
> load on allocation bodies, AARNet already sub-allocates OIDs to
> Australian universities out of its ISO allocated space for use
> in universities' private LDAP schema and any private SNMP MIBs.

I agree that OIDs would in theory be more flexible, since they
include enterprise numbers, but in practise most everyone who will
care about this has an enterprise number anyway, and they are
easy to obtain.  I would rather leave it simple.

> Because of the OID/LDAP linkage, I'd expect DNS registries to
> be the ultimate allocators of OIDs.
> 
> >    ... Since a component of a
> >    DNS name cannot start with a digit, there is no risk of confusing
> >    the two.
> 
> Not so, this requirement was removed some time ago.  Consider
> http://www.3com.com/.

I'll fix my text; I meant to say that the first component (com, org, ...)
cannot start with a digit.  The 3com example is still OK, as John
had shown.

> It would thus be better to list the namespace explicitly rather
> than make any assumptions about DNS names.  Especially as DNS
> naming is going to go through some major changes to allow for
> multilingual names.
> 
> >  - Less transcribable - OUI normally expressed as six-digit hex
> >    number; schemes such as MAC address and EUI-64 are expressed
> >    as 12 to 16-digit numbers.
> 
> OUIs have a OID form, so the OID form should should be either be required
> or forbidden to prevent confusion.

OUIs have been removed as a possibility; anyone wanting to use
an OUI should use the eui. format.

> It probably best to treat the OUI as a 12 or 16 hexdigit number.
> Using just the OUI is problematic for huge organisations, as they
> then need to track iSCSI namespace use internally.  Assigning
> a "MAC address" (OUI + 3 octets) to the iSCSI team is easy
> administratively.
> 
> Finally, we should use the URI name and format for the namespace
> where a URI format exists.  This is simply for consistency.
> 
> For example:
>    backwardsdns:au.edu.example.faculty
>    oid:1.32.43.5.3.2.43.2.2.34
>    oui:2e319c65786e

I had suggested this before, in my draft on iSCSI URNs; the IESG
completely shot this down, and I'm still not sure why.  Anyway,
I don't have the energy to push the URN/URI thing any further.

> Regards,
> Glen
> 
> --
>  Glen Turner                                 Network Engineer
>  (08) 8303 3936      Australian Academic and Research Network
>  glen.turner@aarnet.edu.au          http://www.aarnet.edu.au/
> --
>  The revolution will not be televised, it will be digitised

-- 
Mark A. Bakke
Cisco Systems
mbakke@cisco.com
763.398.1054