Re: [COSE] [Last-Call] Iotdir telechat review of draft-ietf-cose-x509-07

[Ivaylo Petrov <[e-mail]>]

Hi Carsten,

Thank you for this review! I have asked if anyone would be willing to
generate the examples for this draft, but for the time being I have removed
the paragraph about this. As for the discussion about bag vs some other
term, I believe I see your point, but my understanding from this thread [1]
is that bag is a well defined concept and we could still use it here.

[1]: https://mailarchive.ietf.org/arch/msg/cose/qYkms2KEZQbfJtBKRoFnaGU7ILE/

--
Best regards,
Ivaylo

On Tue, Oct 20, 2020 at 1:43 PM Carsten Bormann <[e-mail]> wrote:

> On 2020-10-20, at 03:15, Michael Richardson <[e-mail]> wrote:
> >
> > I believe we use the term bag because it is permissible for a certificate
> > artifact to appear more than once. Stupid maybe, but permissible.
> >
> > I think that some systems/libaries considered the Issuer/Subject to be
> the
> > key for indexing the set, and then they got confused if there was more
> than
> > one certificate in the bag.  The additional object used a different
> signature
> > and/or hash.  At least, I have some dim memory of some situation being
> > described to me.  I think that the names of the guilty parties were
> withheld.
>
> I think we have a different perception of what “is” means.
> In my shopping bag, there *is* a difference between having one or two
> yoghurts in there.
> In the x5bag, having the same certificate twice is exactly equivalent to
> having it once.
> So it “is” a (non-empty) set, not a bag, even if the *representation* (as
> an array, with a special case for the singleton) can actually have
> duplicates.
>
> Given the semantics, the question how one “finds” things in that set is
> more of an implementation question.  I don’t think offering this as a
> multimap(*) with some arbitrarily chosen map key is flexible enough.
> Normally, a simple iterator (so you get to see any and all of the elements)
> will be the best solution, because the implementation cannot know what the
> application-specific validation process is looking for, and we are talking
> about a very small set.
>
> Grüße, Carsten
>
> (*) Cannot be a map, as there is no guarantee of uniqueness of any key.
>
>