|
RE: Activation points and callout modes
2005-01-31 17:25:13
Dear Martin,
in order to be able to better respond you I need to know a few things.
Please do not read this as agressive, I just want to know the vision of
this group. Since I see that I am the only one asking questions, I
apologize if this delays the whole group or if this keeps this mailing list
unecessarily buzy.
- who are the "we" when you say "we won't do that as a humman command
interface" ?
- what is the purpose of this OPES work? Is it just to carry some research
or to design operational network building blocks?
- the language being used "OCP server/client" and an "MTA becoming an OPES
processor" and a "callout server" does not make ANY sense to me. This is a
middlebox language where the middlebox protocols are affected/enhanced by
an external feature. This is not a server based Service language. IMHO this
confusion does not help.
At 13:27 31/01/2005, Martin Stecher wrote:
Hi jfc,
seems that the confusion is ongoing ;-)
I also have problems to get your picture of OPES :-(
Frankly, my feeling is that from the very first begining, we were not able
to work on blackboard together and to sort out what we are talking about,
what is its use, who is interested in using, financing it, running a
business with it. Hence the confusion.
The reason why I quoted you in my posting was because Tony has
wrote in his message from Jan 22: "Following on from what jfcm said, I
can't see any use cases ..."
>
> 1. I set-up Machines to offer OPES. I set them up as a network, using
> hopefully a single protocol to dialog together.
I have already problems with this. "... to offer OPES" - what shall this mean?
In order to offer open pluggable edge device services you set up a network
of machines that communicate which each other. This is what I read.
Yes. As you know a "Service" is something that people want to use for
themselves or on their machines and they will pay with a credit card. (BTW
I do not understand the difference you make between OPES and OPEDS?).
My OPES world is: There is a callout server the same way we know it already
today as an OCP server for HTTP or as an ICAP server;
What is that "callout server = OCP server" going to serve for in your idea
? Is it just a single black box with no back-up or does it belong to a
domain of servicing processes on several machines ?
this now needs to be connected to an existing SMTP structure.
Here I am lost. I understand what is a Service servers structure , like an
-email service mailboxes server, like a web hosting services Apache server,
like a Banking service server, but I do not know what a "small mail
transfer protocol structure" is. To my underestanding there are computer
ports, where datagrams are sent based upon DNS information, where they are
processed by several modules, accomplishing different tasks fitting a
protocol desciption.
We want to achieve this by simply adding an OCP client module to an MTA
which by this then becomes the OPES processor.
If I read it correctly, you want to add an OCP client (what I read as an
communication process able to establish and maintain a cession with another
communication process using the OCP protocol). I frankly do not understand
how this can modify the MTA if in the MTA architecture there is no one or
several filters able to trigger and feed that communication. The important
point is to understand where to put them in real world MTA and to what
purpose. There are different MTA architectures and at least three functions
: receive, store and forward, with a loop store-forward when the forward
fails which makes four processes to consider (at least) receive, loop,
store, forward, and 8 possible places to modify the data fed into the
processes (before and after) and three different types of changes: on the
data, on the header, on the information passed to the function. Now
different MTA have different additional possibilities due to their own
architecture (for example local/external mail, distribution lists, aliases,
etc. etc.)
Also, I do not understand the semantic of "OPES processor". But I
understand the idea of an MTA becoming OPES processed ?
> 2. I install OPES filters/probe, etc. but with a naming that
> a client will
> understand.
> - They will dialog with the above with the same protocol.
> - I will install them as I want and along the clients
> application
> design and the intended service. If I need to have 10 of them
> in an MTA, I
> need to be able to just do that.
Totally lost with this.
10 filtering services in an MTA?
They are in a callout server and not in the MTA. There is only an OCP client.
We are talking of services, not of software modules (nor of protocols). You
can imagine all the filtering strategy you want, at whatever place. Or do
you understand "service" in OPES as "offeriing OCP protocol support"?
> 3. I figure out there are at least two types of probes
> - to filter a data stream (as in http) with data coming by chunks
> - to filter static data with data coming with a complete file.
Agreed. The first is what I desribed in points 1,3,4
The second is 2 in my posting.
Then I may have a simple synthetisis problem. Could the be presented the
following way (I dont ask you do it, just if it makes sense)?
- here are the two modules
- here is what they permit
- here is what neither of them permit.
> 4. Now I need to establish what in the protocol belong to the
> server and to
> the probe, what belongs to communications and what belong to
> processing
> (server and filtering), so I need to know the language I use
> on the probes
> and the way I drive it from the server. My expectation is
> that when I am on
> the server, I can enter "$> ocp" and I have an "$> ftp" like answer,
> including the "help".
We won't do this as a human command like interface.
Probably using again OCP with a profile for SMTP.
The connections will be opened by the OPES processor (OCP client)
not the callout server (OCP server).
Who said so? I want to control the OPESed boxes of my domain.
I hope this helps you to clarify (at least that we do not speak of the same
thing)
jfc
PS. there is an IETF meeting in Paris in August. Do you think you will make
it (or anyone else ?): I could get a room for a full day with a blackboard,
refreshments and snacks - I know it makes a long time, but I can stop
questionning, seeing what comes out of your proposition. We would have a
review by then. I would be a user looking at your product [this is just
what I am: a user who develops his own tools when he cannot find or advise
them].
|
|