Re: [Qirg] Review of draft-irtf-qirg-principles-02

[Wojciech Kozlowski <[e-mail]>]

Hi Mathias,

So, I finally included your feedback. Thanks a lot! I've skipped responses etc., but below you can see which comments I included, which I decided to omit, and which I think need further discussion.

Diff for this changeset: https://github.com/Wojtek242/draft-irtf-qirg-principles/commit/6b3009645e34fe9a844542a98747aed7557b66b9

Latest version: https://github.com/Wojtek242/draft-irtf-qirg-principles/blob/master/draft-irtf-qirg-principles-03.txt

Thanks,
Wojtek


On Mon, 2019-12-09 at 09:48 +0000, VAN DEN BOSSCHE Mathias wrote:
Page/section: p 6/ § 3

Proposal/Question:
I would change the title of §3 to _Entanglement as the
fundamental resource_

Explanation:
The fact is that a service is an action, while a resource
is something you can create, quantify and manage. There is
a targeted service is to "distribute entanglement". BTW
the document mentions entanglement as a resource §§ 4.4.2.1
and 5.6.1, so that would improve consistency.

[WK] Good point - done

Page/section: p 10 / § 4.3 (last 3 lines)


Proposal/Question:

change
The unknown quantum state that was transmitted never entered the
network itself. Therefore, the network needs to only be able to
reliably produce Bell pairs between any two nodes in the network.

to
change
The unknown quantum state that was transmitted was never fed into
network itself. Therefore, the network needs to only be able to
reliably produce Bell pairs between any two nodes in the network,
and perform BSM with external qbits to be transmitted.

Explanation:
* It is a delicate concept to explain that the network is here only
to create a direct end-to-end link, which is then used directly
by the end nodes without having the intermediate nodes to handle it.
* The interface function (the BSM with external qbits) I think is
still part of the network and should not be forgotten

[WK] Done, but skipped mention of BSM, because that would require an explanation of what it is and I don't think it would enough value in this section.

Page/section: p 10 / § 4.4.1

Proposal/Question:

The title could be _4.4.1 _Elementary link generation_ instead.

The expected contribution should include at least 3 main parts
* Create the BPS
* Store the entangled members of non-local BPS
* Monitor the entanglement resource level on each link

Explanation
The other aspects of entanglement life cycle are well described
in the subsequent sections, it looks like this one is dedicated
to elementary links, right?
Everyone will have in mind the function to create BPS across the
elementary links, but we should first not forget the fact that
they need be stored (esp. in the light of the asynchronous use
mentioned further in the document). And then this stored entanglement
will be the resource to be monitored on each elementary link,
because it will be consumed and it will expire (by decoherence).
Monitoring will detect low levels and trigger regeneration.

Can we agree on that before developing further?

[WK] I'm still working on this section and will send an update soon. Everything you mentioned is basically there, but I think monitoring already goes into too much detail and should be left to actual protocol implementations.

Page/section: p 11 / § 4.4.2

Proposal/Question:

Augment

The problem with generating entangled pairs directly across a link is
that its efficiency decreases with its length. Beyond a few 10s of
kms the rate is effectively zero and due to the no-cloning theorem we
cannot simply amplify the signal. The solution is entanglement
swapping.
with

The problem with generating entangled pairs directly across a link is
that its efficiency decreases with its length. Beyond a few 10s of
kms in optical fibre or 1000s of kms in free space [ESA, Micius], the
rate is effectively zero and due to the no-cloning theorem we cannot
simply amplify the signal. The solution to propagate entanglement at
large distances is entanglement swapping.

Explanation:

The two kinds of links (fibre & free space) shall be mentioned for
completeness. They are not competing, rather complementary, and both
will require swapping and routing devices anyway. Free-space to satellites
will allow having fewer nodes on the path, the goal being to improve
capacity.

[WK] Good point - done

Page/section: p 12 / § 4.4.2

Proposal/Question:

Add at the end of the section:
<<
Entanglement swapping allows entanglement to exist at larger distances
than what the entangled degrees of freedom can propagate. Yet one should
keep in mind that the capacity of the network still decreases with the
number of swapping nodes (all other parameters being kept fixed). So
entanglement swapping solves the problem of existence of long distance
entanglement, but not that of performance of the end-to-end link.
>>

Explanation:
This
- clarifies the role of entanglement swapping in the network as a system
- allows identifying different way of attacking the problem of performance.

[WK] This is a good point, but I'm not sure how best to include it. I agree with the point you are making, but to include I would like it to a bit more detailed - why does it reduce capacity? There are various factors to consider in this and it would be great if they could get some coverage - this is definitely still a TODO.

Page/section: p 12 / § 4.4.2.2

Proposal/Question:

Change

2. An identifier that allows the application to unambiguously
determine which qubits at the two end-points belong to which
entangled pair.

to

2. An identifier that allows the applicatqion to unambiguously
determine which qubits at the two end-points belong to which
entangled pair (including end-point identification).

[WK] I've changed the text this is referring to be shorter since the original wasn't adding much value.

Page/section: p 12 / § 4.4.2.2

Proposal/Question:

Add

4. An estimate of the expiry date of the BSP.

Explanation:

It might not be relevant in the first networks where decoherence
will require fast consumption of the created entanglement resource,
but we hope that this constraint will be relaxed, and that this
will become a relevant network management information.

Alternatively create a §4.4.2..3 on entanglement storage

[WK] This is also a good point, but just like with link monitoring I feel this should be left to protocols and can be left out of the current draft. May be worth mentioning in the principles section though maybe when we start working on it?

 Page/section: p 13ff / § 5

Proposal/Question:

From that point on, the document mentions abundantly the term of
'repeaters'. I recommend to give it up and use a more accurate term
such as swapper or router (that are introduced in the document) or
propagator (that's my QFT background).

Explanation

I know the term is pretty much used already, but here is my argument.

Although it is a convenient term for those who have a classical
network background (where it is fairly transparent), it is abusive
in the case of entanglement networks. The fact is that nothing is
repeated, it is just blank entanglement that is forwarded, swapped,
propagated commuted (as in the old telephone systems where an operator
had to plug a wire between two lines nodes to connect two network
segments and allow an end-to-end line to exist). It repeats no data
(as the considerations at the end of the document insist on).

Using the term of repeater is confusing, and leads to wrong understanding
and then inadequate decisions.

[WK] This is a good point. However, I am strongly in favour of using the word quantum repeater in favour of not alienating the academic community that has done all the creative work on this so far. We do have additional terminology now on top of quantum repeaters which might help.

Page/section: p 14 / § 5.1 (3.)

Proposal/Question:

Change

This implies ... store temporary state as the ...

to

This implies ... store the received temporary state as the ...

Explanation:

If I understand the sentence right.

[WK] The temporary state may be received, but it may also be metadata that is only meaningful to the local node (e.g. which interface a qubit was received on).

Page/section: p 18 / § 5.5.1

Proposal/Question:

Change

"The memory lifetime depends on the particular physical setup, but the
highest achievable values currently are on the order of seconds."

to

"The memory lifetime depends on the particular physical setup. This
is object of intense research. Current setups achieve lifetime
up to the order of seconds, but promising approaches should allow a
significant expansion of this value"

Explanation:

I came across a paper [Nature _517_, 177 (2015)]  that claimed
6h coherence time with Europium ion spins that are "optically addressable".
Since I am no experimentalist, I do not know what intricacies this
setup might require. I do not know either whether this result is
object of controversy. But if this is not the case, I would suggest
changing the text to something less hard on memory lifetimes.

[WK] The text already mentions that memory lifetimes will keep improving.

Page/section: p 18 (and 19) / § 5.5.1 (and 5.5.2)

Proposal/Question:

Explain why it might be more cost efficient to use short lifetimes.

Explanation:

It might be a valid choice, but it deserves motivation. Do not
forget that anything expensive produced in mass figures becomes
inexpensive.

[WK] I just removed mention of cost-efficiency - it's just too early to talk about these things. Thanks for pointing it out.

Page/section: p 18 / § 5.5.2

Proposal/Question:

It is not clear what the mentioned success rate of 1e-3 addresses.
I understand it is the rate of success to create a BPS by pumping
a non-linear crystal with a suitable pump. Is it this? Or is it
the probability that BSM effectively entangles 2 BPS? Or is it
something else?

It is tempting with this figure to compute the capacity of an
n-hop path. But is they the correct needed information?

[WK] I meant elementary link generation attempts. I changed this to say "small fraction", because ultimately I was only thinking about one particular architecture for that figure. It would also be tricky to use that number in estimates as it is possible to multiplex these attempts in time and space (current work going on).


Page/section: p 21 / § 5.6.1 (5.)

Proposal/Question:

I propose to remove the considerations on QKD, or at least
to reduce them drastically.

Explanation:

There might be other solutions to secure the control plane other
than QKD. It is true that once optical links are available, it
is tempting to implement QKD protocols, but is it the right moment
to highlight a particular solution?


[WK] Good point, will keep in mind when reworking the principles section (still needs doing). Same applies to all the comments below.

Page/section: p 22 / § 5.6.2 (3.)

Proposal/Question:

Change

"This point is crucial in enabling the reuse of resources of a
network and (...)"

to

"This point is crucial in optimising the use of entanglement
resources of a network and (...)"




Page/section: p 24 / § 5.6.2 (6.)

Proposal/Question:

Question : do you consider parallelising over the same path with
e.g. multiplexed optical channels, or do you consider also
parallelising over different paths? In which case it could be
mentioned.




Page/section: p 24 / § 5.6.2 (7.)

Proposal/Question:

I would suggest giving synchronisation accuracy needs here.
Not necessarily the figure but a least the relationship
with the source rate etc.

Explanation:

This would allow assessing what is feasible and what is not.
At some point, synchronisation may be much more demanding than
what you can do with GNSS (~ 0.1 ns). As a matter of fact, the
network its might become itself a big machine to measure the
metric of space time on Earth....