Could you and others on the PILC list help?
John Border wrote:
> A minor point re the latest ASYM draft...
>
> In Section 6 and 8, the ASYM document makes a comparison to the PEP
> document which is not exactly accurate. It says that the documents differ
> because ASYM limits itself to techniques which do not alter end to end
> semantics while implying that PEP only addresses techniques which do alter end
> to end semantics. However, this is not accurate. The PEP document discusses
> both kinds of techniques. The real difference between the documents is that
> the PEP document, being only a survey document, only touches on the techniques
> and does not describe them in the detail presented in the ASYM document...
Thanks John, Mahesh raised this at the end of the last edit, but
we couldn't quite see how to change the text. The two of us
had a meeting today to discuss this and came up with the starting
text at the bottom of this email.
-------------------------
The problem, as we see it:
The PEP draft defines a PEP as:
"An entity in the network acting on behalf of an end system or user
(with or without the knowledge of the end system or user) in order
to enhance protocol performance."
It does say that ACK manipulation is a PEP measure, to a certain
extent it is true, but not entirely. I do not dispute that it is
USEFUL for PEP to describe these schemes - it IS useful.
I would like to be clear that the ASYM schemes are very different
to split TCP / TCP snooping / Application proxies etc:
In ASYM, the ACK manipulation involves, queuing packets, dropping packets
and scheduling packets - these are normal network operations (e.g. a routing
doing, e.g. diffserv, may do all these).
For ASYM modifications to work efficiently requires looking at TCP headers.
The ASYM schemes are either (i) TCP modifications or are (ii) network
forwarding changes - there is no intention to act "on behalf"
of an end system.
--- Can you advise on some good wording which helps the reader understand and is consistent with PEP.As a starter I suggest the following for Section 6:
PLEASE DO SUGGEST CHANGES or alternate text!
"The techniques proposed in this document do not seek to modify the end to end semantics of the TCP packets exchanged between hosts. They therefore do not inspect/modify any TCP or UDP payload data. They also do not modify the port numbers or addresses of packets. Many of the risks associated with other protocol enhancing measures (e.g. local retransmissions, local acknowledgements, split TCP) do not exist for such schemes. A more detailed description of the issues in providing network protocol enhancement is described in [PEP-ID] which also compares the various approaches."
We could include an abbreviated version at the end of the document.
-----
gorry
------------------------------ http://www.erg.abdn.ac.uk/users/gorry ------------------------------ http://www.erg.abdn.ac.uk/users/gorry
This archive was generated by hypermail 2b29 : Mon Jan 28 2002 - 09:12:26 EST