colibri: e2e admission

[juagargi]

Do the end to end admission.
This PR also introduces some small changes to the app messages and capnp for e2e setup messages.

---

This change is 

[oncilla]

Reviewed 23 of 24 files at r1.
Reviewable status: 23 of 24 files reviewed, 5 unresolved discussions (waiting on @juagargi and @oncilla)

---

go/cs/reservation/request.go, line 44 at r1 (raw file):

// IsHopByHop returns true. If a subtype needs not to be hop by hop, it can override the function.
func (m *RequestMetadata) IsHopByHop() bool {

What requests are planned to not be hop by hop?

---

go/cs/reservation/e2e/request.go, line 57 at r1 (raw file):

// SetupRequest represents all possible e2e setup requests.
type SetupRequest interface {
	IsThisASTheSrc() bool

The names are not that great.

Also, this could just be Location() PathLocation

with

type PathLocation int
const (
    UnknownPathLocation PathLocation = iota
    Source
    Transit
    Transfer
    Destination
)

(or similar)

---

go/cs/reservation/e2e/request.go, line 71 at r1 (raw file):

	RequestedBW              reservation.BWCls
	AllocationTrail          []reservation.BWCls
	totalAScount             int

totalASCount (could also be a function that operates on SegmentResvASCount)

---

go/cs/reservationstore/store.go, line 410 at r1 (raw file):

	// admitted so far
	// TODO(juagargi) update token here

Will you address that TODO as part of this PR?

---

go/cs/reservationstore/store.go, line 422 at r1 (raw file):

	var msg base.MessageWithPath
	if req.IsThisASTheDst() {
		asAResponse := failedResponse.(*e2e.ResponseSetupFailure)

Why is this done?

[juagargi]

Reviewable status: 23 of 24 files reviewed, 5 unresolved discussions (waiting on @oncilla)

---

go/cs/reservation/request.go, line 44 at r1 (raw file):

Previously, Oncilla (Dominik Roos) wrote…

What requests are planned to not be hop by hop?

Function removed: if the new header is approved, all control messages will be hopping from COLIBRI service to the next.

---

go/cs/reservation/e2e/request.go, line 57 at r1 (raw file):

Previously, Oncilla (Dominik Roos) wrote…

The names are not that great.

Also, this could just be Location() PathLocation

with

type PathLocation int
const (
    UnknownPathLocation PathLocation = iota
    Source
    Transit
    Transfer
    Destination
)

(or similar)

Done. The enum is slightly different since Transfer is not a location, and a transit node may be also a transfer one. The function for checking if it's a transfer node is now called Transfer()

---

go/cs/reservation/e2e/request.go, line 71 at r1 (raw file):

Previously, Oncilla (Dominik Roos) wrote…

totalASCount (could also be a function that operates on SegmentResvASCount)

Done. The value is pre-computed on construction to allow cheap calls to get it.

---

go/cs/reservationstore/store.go, line 410 at r1 (raw file):

Previously, Oncilla (Dominik Roos) wrote…

Will you address that TODO as part of this PR?

The token needs to be updated also in the segment admission. I want to soon open a PR with the changes to handle the MACs and token, and take care of both TODOs.

---

go/cs/reservationstore/store.go, line 422 at r1 (raw file):

Previously, Oncilla (Dominik Roos) wrote…

Why is this done?

Done.

[oncilla]

Reviewed 1 of 24 files at r1, 4 of 5 files at r2.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @juagargi)

---

go/cs/reservation/e2e/request.go, line 71 at r1 (raw file):

Previously, juagargi (Juan A. Garcia Pardo) wrote…

Done. The value is pre-computed on construction to allow cheap calls to get it.

right. Although there is much more optimization potential elsewhere :)

---

go/cs/reservation/e2e/request.go, line 122 at r2 (raw file):

}

type PathLocation int

consider adding a string method

---

go/cs/reservation/e2e/request.go, line 125 at r2 (raw file):

const (
	Source PathLocation = iota

It would be best to have the zero value be unknown, such that bugs are easier to detect.
Like this, when you forget to set the PathLocation somewhere, it looks valid.

---

go/cs/reservationstore/store.go, line 422 at r1 (raw file):

Previously, juagargi (Juan A. Garcia Pardo) wrote…

Done.

No, I don't understand the intuition behind it. Why do we use the response for failure here?

[juagargi]

Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @oncilla)

---

go/cs/reservation/e2e/request.go, line 125 at r2 (raw file):

Previously, Oncilla (Dominik Roos) wrote…

It would be best to have the zero value be unknown, such that bugs are easier to detect.
Like this, when you forget to set the PathLocation somewhere, it looks valid.

I though about it, but the location is a derived value, it only depends on len(r.AllocationTrail) and it always has a value, i.e. it can never be unknown.

---

go/cs/reservationstore/store.go, line 422 at r1 (raw file):

Previously, Oncilla (Dominik Roos) wrote…

No, I don't understand the intuition behind it. Why do we use the response for failure here?

You are right: the failed response has to be transformed to a success response.

[oncilla]

Reviewed 2 of 2 files at r3.
Reviewable status: complete! all files reviewed, all discussions resolved

[oncilla]

/rebase

[oncilla]

/rebase

[oncilla]

/rebase