Re: Scope of the Web: what protocols are allowed?

On Nov 5, 2003, at 10:55 AM, noah_mendelsohn@us.ibm.com wrote:

Example 1:

> I'll call this mythical system "spread spectrum peer to peer".

...

>    I'm not asking whether this
> is a good way to build a system.  I'm merely asking what the 
> architecture
> document would say about it if I did.

I don't see any problems here - if there's something in the webarch 
draft that could be used to condemn this, I'd think that would be a 
bug.  I suspect that if you were looking in from outside, the behavior 
of this system would be similar to that of freenet BTW.

> Example 2: Less RESTful models
> ==============================
>
> This example is probably less architecturally radical.

I don't see anything wrong with this one, although I'd be inclined to 
use an http URI to point to the thing and rely on media-type 
dispatching to arrange for your magic two-way-streaming software to 
take over, RealAudio kind of does this in that they have these .ram 
files that you get with HTTP that are actually pointers to the 
underlying secret realaudio magic, it seems to work pretty well.  Once 
again, if there's something in webarch that gets in the way we should 
look at it carefully.

> My reading of the current editors' draft is that it comes close to
> allowing for them, but there are also statements such as [2]:
>
>         "Web agents communicate complete or
>          partial information about the state
>          of a resource through
>          representations. "
>
> ...that could be taken to preclude either or both of the examples 
> above.

Hmmm, what comes out of the system when you dereference a URI is a 
representation; I don't think Webarch constrains what gears grind away 
inside the system to squeeze out the representation.  If you think it 
does, Noah, perhaps you could suggest an amendment? -Tim


Cheers, Tim Bray  http://www.tbray.org/ongoing/

Received on Thursday, 6 November 2003 10:57:49 UTC