Re: Non-order processing in persistent connections

Hi,

Zhou Kang wrote:
> In fact, we can make some minor alteration to http server's
> performance. For example, we can add a header to the messages to
> indicate the request/response pair, something like 'Request-ID'. A
> client may create a unique digital ID for every request to the same
> server ,  then the server can return responses with this ID or without
> but use the same order. To name this mechanism , I call it as
> 'non-order processing'.

I work on a project for anonymous WWW access. It will base on HTTP/1.1
If you send a *big* packet through a MIX-Net (like Mixmaster, can
only work with a constant block size) a Cache-Proxy must connect the
smaller
parts. You can use the Content-Range Field but if some anonymous users
use
the same Range than they will get wasted data.
So we need also a Message-ID or 'Request-ID'.
This Message-ID should be defined for Request and Response and
*randomly* choose by the client/server. If a server, (user?),
proxy or MIX detect a collision
than a warning or error can be response to the client (may be
server?).   

Bye
Uwe Danz
(my English is too bad: same mistakes corrected)

Received on Monday, 18 May 1998 08:31:24 UTC