So...

One of the things I have been trying to understand over the last couple of
weeks is what entities we have to name in TAPS, and how we give the
entities names that seem to be understood by our community and preferably
are common with other uses in the IETF.

I came across one term that appears to be defined differently in the
charter discussions, to how I would normally understand this. This email
tries to explain this, if you think the same as me, please speak-up and
say this, if you think another definition is better, please say why.

I think the group just needs to decide at this stage, so that all actual
RFCs that are made will be consistent.

---
1) Transport Service

I see a "Transport Service" as an end-to-end service that has
traditionally been mapped by transport people  down to a "port/socket"
(perhaps the "S" part of tsv).

This appears consistent with other uses, I have come across. However, a
service is usually composed of a number of "components" - I seem to recall
these be called "service elements" in OSI. Examples are: reliability,
multicast, confirmed delivery, congestion control, integrity, minimal
latency, etc. (we could, and probably should make a list).

Application
   |
Transport Service(s)
   |                  |
Service          Service
Component   Component(s)

Hence, I see a service as offered via an API to an App or upper-layer
protocol. I see a service as having multiple components. This is not
necessarily the usage in the final charter fro TAPS!!

---

2) When the charter says; "Four examples of Transport Services are reliable
delivery, ordered delivery of data, content privacy to in-path
devices, and minimal latency."

- I think these could indeed be transport services, but I'd actually now
suggest they could be better described as components/elements of a
transport service. In my thinking, one could always build an entire
service that was oriented at "minimal latency" - but equally one could
conceive of this as a transport service component that could be combined
with security, ordering, etc. (using the language above).

---

3) Transport Protocols

Transport protocols are not a service, because many transport protocols
can provide multiple services (again my own terminology). Indeed a single
transport can negotiate to support multiple "features". A "feature"
represents a choice of protocol mechanisms/parameters to instantiate a
service (or one component of a service, e.g. use of ECN, optional
integrity
checks, etc). This term has been sued before, one of the older uses could
be FTP feature negotiation. DCCP formally included "features" in its
protocol design.

Transport protocol
   |                  |
Feature          Feature(s)
   |
Protocol
mechanism(s)

---

For anyone in fear this is endless, it is NOT going to be. Mirja will
present slides to ask for confirmation of the hierarchy terms at the
meeting next week, and after a <shot> discussion we plan to very quickly
and publish a new revised ID with the WG-"approved" terms and use these
hence-forth.

Gorry


_______________________________________________
Taps mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/taps

Reply via email to