W3C home > Mailing lists > Public > www-tag@w3.org > March 2007

Re: Clean layer model of the internet?

From: Jon Hanna <jon@hackcraft.net>
Date: Fri, 09 Mar 2007 15:34:11 +0000
Message-ID: <45F17E73.6080808@hackcraft.net>
To: Jacek Kopecky <jacek.kopecky@deri.org>
CC: Chris Wilper <cwilper@cs.cornell.edu>, Tim Berners-Lee <timbl@w3.org>, www-tag@w3.org

Jacek Kopecky wrote:
> You say "application protocol" layer is cleanly defined in the IP stack,
> and then you say that you can include SSL under another protocol (e.g.
> under HTTP to get HTTPS). But then, on which layer is SSL? It's built on
> top of TCP, but used under HTTPS just like TCP is used under HTTP. And
> HTTPs is on the same layer as HTTP, I assume.

I don't think that's a problem. If an application protocol acts like a 
transport protocol to what is above it then I don't think anything's 
broken. If SSL wasn't designed to be used as a transport protocol by 
what is above it, then that would be different.

> So it seems to me that the two properties above collide in their
> value2web fields, and frankly, I'd drop the first property, "clean layer
> model", because I don't think it is. It may be a "simple layered model",
> but I think that's the point of the second property.

That HTTP doesn't need to know that it is sitting on top of SSL, TLS, 
TCP or something invented tomorrow seems like a clean model to me.

The fact that HTTPS is defined and named differently I don't think is 
clean. (IIRC, it would now be against IESG policy to give it a different 
port number to HTTP and I understand [though my knowledge here isn't 
great, and I'll be dropping out of this thread once it's gone past 
blue-sky suggestions] that this isn't unrelated to the lack of 
cleanliness here).
Received on Friday, 9 March 2007 15:35:23 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 15:32:51 UTC