Re: Mini WGLC draft-ietf-multi6-multihoming-threats-02.txt

[Iljitsch van Beijnum <iljitsch@muada.com>]

On 1-dec-04, at 16:35, Brian E Carpenter wrote:

> > > "Identifier" is defined very differently from the use of "ULID" in 
> > > the more recent DT drafts, to the degree that an ULID can't be an 
> > > identifier according to this document's definition.

> > I think that this is a good point...
> > I am not sure that all of the threats related to redirection exist 
> > when you use ULIDs vs. a pure ID/Loc split.

Not if we require initial connectivity to happen using the ULID as a 
reachable address/locator.

(But I think we'll want to be able to repair an initially unreachable 
ULID using alternative locators in the future, so we probably shouldn't 
depend on this advantage more than we have to.)

> > Are there other places in the document where the threat model would 
> > be different for ULIDs than for IDs that are not also usable as 
> > locators?

Hard to say. However, the threat model is radically different when HBA 
(or CGA) is in use, so a document that looks at this will be very 
different from this general purpose document. I'm not sure if analyzing 
ULIDs without HBA/CGA is useful, considering the proposed solutions 
that are on the table.

> Personal opinion: this document is intended to discuss generic threats,
> and I think it's a bit unfair to expect it to discuss threats for a
> model that hadn't even been invented when the document was almost 
> final.

I agree. My problem with the identifier definition is that an ULID 
doesn't fit this definition, most notably because:

                    The identifiers are not associated with an
                    interface.

My conclusion is that the definition is too strict so it should be 
changed to encompass ULIDs so nothing stands in the way of applying 
this document to solutions that use ULIDs.