The MPLS WG Archive

["Adrian Farrel" <afarrel@movaz.com>]

Tom,

This comes close to my question a little while back.
The issue is not "How does the management application know what it is
provisioning?", but "How does the device know what is being provisioned?"

If you are saying - it knows it through some proprietary MIB field that people
will add to the their own MIBs - this is fine, but I say let's add an explicit
applicationIndex field that implementations can interpret freely.

If you are saying - the application index is embedded in the inSegementIndex - I
say this doesn't work because there is no way of communicating the requirements
to the getNext object on a GET command.

Adrian

----- Original Message -----
From: "Thomas D. Nadeau" <tnadeau@cisco.com>
To: "'Wijnen, Bert (Bert)'" <bwijnen@lucent.com>; "'MPLS WG'" <mpls@UU.NET>
Sent: Tuesday, June 10, 2003 3:59 PM
Subject: RE: prequeal to WG lat call om the LSR mib module


>
>
> > -----Original Message-----
> > From: Wijnen, Bert (Bert) [mailto:bwijnen@lucent.com]
> > Sent: Tuesday, June 10, 2003 3:46 PM
> > To: tnadeau@cisco.com; 'MPLS WG'
> > Subject: RE: prequeal to WG lat call om the LSR mib module
> >
> >
> > So if you have 3 application managing Label Space as
> > in the following example.
> > >
> > > Application Incoming Labels Owned by Application
> > > LDP { 15, 29, 45 }
> > > VPN { 30, 31, 32 }
> > > TE  { 16, 17, 18 }
> > >
> > And then you have the object mplsInSegmentIndexNext
> >
> > How does that IndexNext object provide me with a proper
> > new index value that I can use. Cause it does not know
> > if I (as an application) want to create a new InSegment
> > for LPD, VPN or TE... or does it? If it does, then I
> > do not understand how. Maybe I missed the explanation in
> > some text in the MIB document... if so, pls point me to it.
>
> It knows because it knows which device it is
> provisioning and thus knows what the application ID
> values need to be (from its agent cap statement or
> other documentation). I don't see this as a problem since
> there is a whole slew of non-standard things that a
> provisioning system needs to know about devices it
> manages. In fact, I would bet that %100 of the implementations
> out there will require some bit of proprietary MIB support
> to facilitate provisioning of static LSPs via SNMP (if
> they even will allow this). IMHO the fact that we have %90
> of what most people need is not too bad.
>
> --Tom
>
>