Re: Weekly posting summary for [e-mail]

[John Day <[e-mail]>]

Yes, "association" was an application level term in OSI.

This was because the wire stringers, i.e. PTT types, had forced a 
definition of "connection" that was wrong.  In OSI, a "connection" 
was between the (N+1)-entities (protocol machines).  Not the 
(N)-entities as it should be.  TCP gets this wrong as well, where 
connections are between port-ids that are at the layer boundary. 
(Connection here is used to be interchangeable with flow.  Not 
intended to imply only those protocols with feedback mechanisms.)

This error leads to all sorts of problems, which are avoided by 
delta-t which decouples port allocation and synchronization.

The convention of the use of port-ids is to create a "connection-id," 
i.e. to distinguish multiple flows between the same two addresses. 
Early on there was considerable experimentation with how to do this. 
Some protocols would have a "connection-id" field and one end would 
number from one end of the field and the other end from the other 
end.  That of course had problems.  Since port-ids were locally 
unique, the idea was hit upon that concatenating them would be a 
simpler way to do it.

The kludge of using them for application names came later.

It turns out the port-id concept is crucial to getting layer boundaries right.

This discussion has only touched the tip of the iceberg on the 
analogy of protocol and program.  There is the Protocol as Program in 
the large, i.e.  across all systems, e.g. the TCP spec laid out in 
the RFC.  There is the protocol in a system in a layer (one might 
have multiple of these for different security domains or some 
such).and there are then multiple instances of that for each 
flow/connection.

As to what is a protocol, someone already hit pretty close.  a set of 
rules and formats (semantic and syntatic) which determines the 
communication behavior of a state machine. 

It has been amusing how often some think that merely writing down the 
format of the messages is a protocol specification.

Take care,
John




At 0:53 -0500 2012/01/06, Thomas Narten wrote:
>Total of 59 messages in the last 7 days.
>
>script run at: Fri Jan  6 00:53:02 EST 2012
>
>     Messages   |      Bytes        | Who
>--------+------+--------+----------+------------------------
>  11.86% |    7 | 12.60% |    57079 | [e-mail]
>   8.47% |    5 |  8.91% |    40337 | [e-mail]
>   8.47% |    5 |  7.46% |    33795 | [e-mail]
>   8.47% |    5 |  6.79% |    30739 | [e-mail]
>   6.78% |    4 |  6.67% |    30218 | [e-mail]
>   3.39% |    2 |  6.98% |    31588 | [e-mail]
>   3.39% |    2 |  3.08% |    13944 | [e-mail]
>   3.39% |    2 |  2.95% |    13375 | [e-mail]
>   3.39% |    2 |  2.83% |    12836 | [e-mail]
>   1.69% |    1 |  3.24% |    14685 | [e-mail]
>   1.69% |    1 |  2.65% |    11995 | [e-mail]
>   1.69% |    1 |  2.42% |    10981 | [e-mail]
>   1.69% |    1 |  2.30% |    10413 | [e-mail]
>   1.69% |    1 |  2.00% |     9065 | [e-mail]
>   1.69% |    1 |  1.98% |     8972 | [e-mail]
>   1.69% |    1 |  1.92% |     8676 | [e-mail]
>   1.69% |    1 |  1.67% |     7580 | [e-mail]
>   1.69% |    1 |  1.65% |     7474 | [e-mail]
>   1.69% |    1 |  1.55% |     7001 | [e-mail]
>   1.69% |    1 |  1.54% |     6984 | [e-mail]
>   1.69% |    1 |  1.50% |     6770 | [e-mail]
>   1.69% |    1 |  1.46% |     6629 | [e-mail]
>   1.69% |    1 |  1.44% |     6522 | [e-mail]
>   1.69% |    1 |  1.44% |     6502 | [e-mail]
>   1.69% |    1 |  1.41% |     6390 | [e-mail]
>   1.69% |    1 |  1.41% |     6365 | [e-mail]
>   1.69% |    1 |  1.41% |     6364 | [e-mail]
>   1.69% |    1 |  1.35% |     6118 | [e-mail]
>   1.69% |    1 |  1.35% |     6105 | [e-mail]
>   1.69% |    1 |  1.31% |     5944 | [e-mail]
>   1.69% |    1 |  1.29% |     5848 | [e-mail]
>   1.69% |    1 |  1.21% |     5495 | [e-mail]
>   1.69% |    1 |  1.18% |     5356 | [e-mail]
>   1.69% |    1 |  1.03% |     4685 | [e-mail]
>--------+------+--------+----------+------------------------
>100.00% |   59 |100.00% |   452830 | Total
>
>_______________________________________________
>Ietf mailing list
>[e-mail]
>https://www.ietf.org/mailman/listinfo/ietf