[Tsvwg] Re: [e2e] e2e principle..where??....
day at std.com
Sun Jun 3 14:07:48 PDT 2001
At 12:40 -0400 6/3/01, Manish Karir wrote:
>yes, but whose semantics are we talking about, I think is what the
>original question was trying to get at.
>The e2e principle as outlined in the paper..is quite general (like saying
>"thou shalt not lie") and I was attempting to figure out its
>boundaries by applying it to a very common and simple example of
>so getting back to the example then, of client---proxy----server
>the client--proxy segment sees "correct" semantics...
>the proxy--server segment sees "correct" semantics...
>the proxy does not tinker with the contents
>so can we argue that this system does not violate e2e?
I believe you are correct. This does not violate the principle. As
has been pointed out, the argument was originally used to determine
what functions went in the network and which in the host. The
earliest use of it I remember were in the Louis Pouzin's papers on
Cyclades that argued for a pure connectionless network (which
acutally preceded the connectionless support in the ARPANet by a few
years.) I believe that the paper even alludes to it being a specific
application of a more general principle OR maybe the same "kind" of
thing shows up in other places. But basically, some things HAVE to
be done in the host and some things HAVE to be done in the net and
there are other things that don't have to be done in either place and
the question is where? The primary argument at the time was: the
PTTs arguing that you didn't need transport protocols that hop-by-hop
was reliable. While the networking crowd argued that hop-by-hop was
not reliable and never could be and regardless no host was going to
trust the network anyway, so if the hosts were going to do error
recovery, then the network didn't have to do as much and could be
simpler and cheaper. The PTTs didn't and don't like this because it
makes the network a commodity and a commodity business is hard work.
(Notice how some large router vendors have learned this and talk
about the importance of putting more intelligence in the network.
They don't want to work hard for a buck (or don't know what to do)
and are trying to turn the Net into the phone company model.
Since the end2end principle is a statement of what is in the hosts
and what is in the network, everything that is in the application
layer is in a host. As someone else said, anything that intercepts
and interprets data intended for layers at transport and above
violates e2e. As I think the same person said, if the intermediate
looks at the higher layer, it must do something about it (whether it
modifies the data or not) otherwise why look at it? Now this sort of
definition could include the @home type proxies. Although, I would
argue that since one might get different results depending on whether
the proxy was used, it is a violation. Others might argue that if
the user didn't know the difference, it isn't!
More information about the end2end-interest