Hello all, During the last OGF we have had very productive discussions, and I think we have made quite some progression. However, one thing that we were not able to agree on was terminology for links/circuits/connections/tubes. Enclosed with this email are two figures that I created based on terminology and figures in ITU-T G.805[1]. If you're interested in G.805, Freek has written a very readable report on it[2], and we have also written an article about modelling multi-layer networks[3]. Figure 1 shows several terms: - *connection point* is any point that can make a connection (the circles in the figures) - *network connection* is the terminated connection from A to B. - *link connection* is a direct connection between two connection points. - *subnetwork connection* is a connection between two connection points that you can not or do not want to describe the details of. - *tandem connection* is a concatenation of link connections Figure 2 shows a multi-layer network connection between X and Y. If we start from X, we first encounter an *adaptation* to a lower layer. Then we go through another link connection. At that connection point we use *multiplexing* to adapt to a lower layer to connection point A. Then we use the same network connection as in figure 1 to go to B, where we go through *multiplexing*, a *link connection*, and an *adaptation* to arrive at Y. Note that the multiplexed connection can also be viewed as a link connection on a higher layer. The same goes for the connection between X and Y, which is a link connection at the highest layer. And because the connection is therminated there, it is also a network connection. Please share your thoughts on this terminology! Jeroen. [1]: ITU G.805 standard: http://www.itu.int/rec/T-REC-G.805/ [2]: Introduction to ITU-T G.805, Freek Dijkstra et al.: http://staff.science.uva.nl/~fdijkstr/publications/G805-introduction.pdf [3]: A Multi-Layer Network Model Based on ITU-T G.805, Freek Dijkstra et al.: http://staff.science.uva.nl/~fdijkstr/publications/G805_Multilayer_Model.pdf -- My email address has changed to <vdham@uva.nl> (The science has disappeared from my address, but I'm still doing it)
My first mail may have had a slightly confusing topic, but I'd like to draw attention to this again. At the last OGF we had a continuously recurring discussion about what to name things (connections, links, ports, interfaces). In the previous mail I've proposed some terminology, please let me know what you think about this. Jeroen.
Jeroen van der Ham wrote:
I'd like to draw attention to this again. At the last OGF we had a continuously recurring discussion about what to name things (connections, links, ports, interfaces).
In the previous mail I've proposed some terminology, please let me know what you think about this.
Not surprising, I like it. First of all, it is an established model in the ITU standardization organisation, and I dislike reinventing the wheel. That said, I have to admit I did not use the G805 terminology all the time. For example, I rather use "port", "interface" or "logical interface" than "connection point" to name a prominent term. When Aaron is talking about "ports", I try to call them "ports" too, while I sometimes let "Interface" slip in (since that is the term we use in NDL). I very rarely use "connection point" in my mails. I've noticed lately two terminology confusions: me and Jeroen about "terminating devices" (I meant edge device, Jeroen meant end hosts). Furthermore, "node" is used for at least network, device and nod in a graph. Sometime we will argue over a term -- after all this is what this WG is about. However, it seems like a waste to do that for terms already established in a decent framework. In the end, the exact name is irrelevant if the concept is clear to all of us on the list. Perhaps I could start my mails with:
In this mail, interface = connection point [1] [1] http://forge.ogf.org/sf/wiki/do/viewPage/projects.nml-wg/wiki/Interfaces
Now, the main question is: Would this be clear to everyone on the list (including the list lurkers), or only more confusing? Regards, Freek
Jeroen van der Ham wrote:
Figure 1 shows several terms: - *connection point* is any point that can make a connection (the circles in the figures) - *network connection* is the terminated connection from A to B. - *link connection* is a direct connection between two connection points. - *subnetwork connection* is a connection between two connection points that you can not or do not want to describe the details of. - *tandem connection* is a concatenation of link connections
Figure 2 shows a multi-layer network connection between X and Y. If we start from X, we first encounter an *adaptation* to a lower layer. Then we go through another link connection. At that connection point we use *multiplexing* to adapt to a lower layer to connection point A. Then we use the same network connection as in figure 1 to go to B, where we go through *multiplexing*, a *link connection*, and an *adaptation* to arrive at Y. Note that the multiplexed connection can also be viewed as a link connection on a higher layer. The same goes for the connection between X and Y, which is a link connection at the highest layer. And because the connection is therminated there, it is also a network connection.
Both a network and a link connection seem to be 'connections between two points'. You say that in one case it's terminated and in one case it's not. Is 'terminated' a property of the link or is it a property of the connection point (or the node the connection point is attached to, depending on one's definition of connection point)? What differentiates a link whose contents gets demux'd to an higher layer vs. a link whose contents get routed at the same layer? A subnetwork connection is a link or network connection that you view as opaque, correct? I.e. I've got a connection between these two points, i either don't know or don't care how it's constructed. If so, would "opaque connection" be a more clear description? A tandem connection is a network or link connection that spans multiple connections? How does it differ from a network connection which in figure 1 also spans multiple connections? Cheers, Aaron
Aaron Brown wrote:
Both a network and a link connection seem to be 'connections between two points'. You say that in one case it's terminated and in one case it's not. Is 'terminated' a property of the link or is it a property of the connection point
It's a 'property' of the connection point -- basically it says that connection is terminated there (these connection points are end-points for a certain layer and are called "termination connection points" or "TCP").
A tandem connection is a network or link connection that spans multiple connections? How does it differ from a network connection which in figure 1 also spans multiple connections?
A network connection is a special case of a tandem connection: it is a tandem connection between termin
What differentiates a link whose contents gets demux'd to an higher layer vs. a link whose contents get routed at the same layer?
I can't answer the question if I were to follow G.805 to the letter. To be annoyingly precise (which G805 is): First: in G805 the term "link" is only used for the physical wire/fibre. What we are talking about is a "transport entity across a link", which is called a "link connection". Second: link connections are never routed or switched. Easiest is to compare link connections and tandem connection with graph theory: - link connection = single edge - tandem connection = path (a sequence of edges) you seem to talk about tandem connections (paths) rather than about link connections (links). So I take it that you mean to ask:
What differentiates a tandem connection whose contents gets demux'd to an higher layer vs. a tandem connection whose contents get routed at the same layer?
The difference is that a tandem connection that gets demuxed to a higher layer must thus be terminated at the lower layer. Thus, that is a terminated tandem connection, which is called a "network connection". (to be even more pedantically correct: the end-to-end connection before termination is called a "network connection", while after termination it is called a "trail")
A subnetwork connection is a link or network connection that you view as opaque, correct? I.e. I've got a connection between these two points, i either don't know or don't care how it's constructed. If so, would "opaque connection" be a more clear description?
Sure -- I'm certainly not saying that I'm a big fan of every word G805 come up with. However, there is a slight difference: a subnetwork connection is a opaque connections *through a given network*. To recap: * link connection is a direct (or atomic) connection on a layer (think: link). * tandem connection is a sequence of link connections (think: a path or a segment of a path) * network connection is a tandem connection that will be terminated (think: a end-to-end path) And for the real G.805-adepts: * trail is a terminated network connection (think: end-to-end path) Regards, Freek -- Disclaimer: This is an e-mail message. Use your own judgment about its value. If you do not have such common sense (e.g. you are a lawyer) or like to see crap like warranties, intended-audience or as-is statements, then the following applies: you do not understand the concept of satire and are not allowed to read this e-mail.
Freek Dijkstra wrote:
A subnetwork connection is a link or network connection that you view as opaque, correct? I.e. I've got a connection between these two points, i either don't know or don't care how it's constructed. If so, would "opaque connection" be a more clear description?
Sure -- I'm certainly not saying that I'm a big fan of every word G805 come up with. However, there is a slight difference: a subnetwork connection is a opaque connections *through a given network*. What is 'network' in this context?
To recap: * link connection is a direct (or atomic) connection on a layer (think: link). * tandem connection is a sequence of link connections (think: a path or a segment of a path) * network connection is a tandem connection that will be terminated (think: a end-to-end path) Ok, the only issue I have is the differentiation between a tandem and network connection. They both seem to be paths, and the differentiating factor is what happens to the data after it leaves the connection (does it get forwarded on at the same layer, or does it get demux'd or adapted up to a higher layer).
Cheers, Aaron
Aaron Brown wrote:
Freek Dijkstra wrote:
[subnetwork] so, would "opaque connection" be a more clear description?
there is a slight difference: a subnetwork connection is a opaque connections *through a given network*. What is 'network' in this context?
A subnetwork as defined by G.805. G.805 defines this in a way that would make Vogon poetry mild by comparison:
3.33 subnetwork: A topological component used to effect routing of a specific characteristic information.
3.34 subnetwork connection: A "transport entity" that transfers information across a subnetwork, it is formed by the association of "ports" on the boundary of the subnetwork.
All that is worth remembering of this is that a subnetwork is part of a layer network, and only contains subnetwork connections of a given layer, and does not give details about underlying layers.
Ok, the only issue I have is the differentiation between a tandem and network connection. They both seem to be paths, and the differentiating factor is what happens to the data after it leaves the connection (does it get forwarded on at the same layer, or does it get demux'd or adapted up to a higher layer).
Correct. It may not be necessary to distinguish between the two in whatever schema NML comes up with. For me it is just a convenience to differentiate between a "path" and a "end-to-end path" in discussions on this list. Regards, Freek
(I sent this reply to Aaron directly by mistake, here's a copy for the rest) Aaron Brown wrote:
Both a network and a link connection seem to be 'connections between two points'. You say that in one case it's terminated and in one case it's not.
Correct. But what I also said is that a link connection is a *direct* connection between two end-points.
Is 'terminated' a property of the link or is it a property of the connection point (or the node the connection point is attached to, depending on one's definition of connection point)? What differentiates a link whose contents gets demux'd to an higher layer vs. a link whose contents get routed at the same layer?
Terminated is a property of the connection between A and B. You can only establish whether a connection is terminated if you know the details of the whole connection, and the adaptations and the deadaptations cancel eachother out. Some of these details of the connection may of course be abstracted through subnetwork connections.
A subnetwork connection is a link or network connection that you view as opaque, correct? I.e. I've got a connection between these two points, i either don't know or don't care how it's constructed. If so, would "opaque connection" be a more clear description?
That is correct, I don't have a preference for either term.
A tandem connection is a network or link connection that spans multiple connections? How does it differ from a network connection which in figure 1 also spans multiple connections?
A tandem connection does not have to be terminated. Jeroen. -- My email address has changed to <vdham@uva.nl> (The science has disappeared from my address, but I'm still doing it)
participants (3)
-
Aaron Brown -
Freek Dijkstra -
Jeroen van der Ham