ietf-openproxy
[Top] [All Lists]

Re: is SMTP a candidate for OPES ?

2004-07-06 11:10:17

At 18:15 06/07/04, Alex Rousskov wrote:
jfc,
        IMHO, the stupidity of judges or laws should not limit OPES
scope. OPES framework (in a broad sense) is applicable to any
communication between content producer and content consumer. Whether
the unit of communication is "stored and forwarded" or "just
forwarded" is irrelevant from architecture/scope point of view. It may
affect how and when the adaptations are performed, but not whether
they can be performed in an OPES-compliant way.

        The old OPES architecture draft may not reflect this and other
scope concerns, but that's a different question/problem.

hmmmm. I am afraid you missed my point. This is what one can name "digital decoherence" like for quantum physic. In the network information is under data state (bytes). Outside of the network (when you see it, use it) it is under object state (a mail, a graphic, etc.). What the Judge says is: data streams are protected by law, but that objects are subject to objects law. Like quanta are subject to quantum law and apples are subject to Newton law.

Decoherence is when an apple electron is both in quantum and newtonian state. Digital decoherence is the same when data information becomes object infomation - being both at the same time. This occurs at the proxy level. So one cas say that OPES philosopy is to take the information as data, to massage it as part of a possible object and tro give it back as data which will later resolve into another object (or several) than initially foreseen. This introduces an interesting case of the layer analysis of what I name the interapplication layer. But the point is not here?

But the point again is : you do not know what a mail is until you have completely stored it. Let say the signature is at the end. Only then you fully know what you want to do with the header. This builds a window of the size of the mail. So two possibilites :

- either you are not modifying something on the fly. You massage a stored file (and you introduce a delay in waiting to massage it to have received it - what I suppose SMTP does not do).

- or you have to find a way to race the network faster than SMTP to tell a filter which has not yet rceived the first datagram for the mail :"the datagram to come is part of a mail opf which the last datagram says that the first datagram should be processed".

I suppose there is a way to do that in doing both ? In using two filters ?
Filter A reads the mail until the end and blocks the last datagram if thre is something to do. Filter B is then told what is to be done and starts processing datagram in seqence until it get one before the last. It signals A to release the last datgram and processes it on the fly. Quite complex ?
jfc


<Prev in Thread] Current Thread [Next in Thread>