Why no one goes further and proposes the  concrete schema and classes for doing
this?

I really have no great knowledge in the area, so the following is just a some
kind of expectation what we really need to do while elaborating such a system. 

Let we play with HTTP protocol and we have just the only EVENT server that
exposes the following REST interfaces:

1) http://newbie.com/feeds lists the HTML docs with the following information 
    <div class="feeds">
     <ul>
      <li class="feed">
       <a href="http://newbie.com/feed/msf-17";>Medical Sensor Feed # 17</a>
      </li>
      <li class="feed">
         <a href="http://newbie.com/feed/msf-27";>Radio Sensor Feed # 27</a>
      </li>
       <!-- etc -->
     </ul>
   </div>

2)  http://newbie.com/feed/msf-11 lists the stream 
     <div class="feed">
           This is a feed from <span class="source"><a
href="http://newbie.com/feeds";>source feed list</a> </span>   with the name
<span class="name">http://newbie.com/feed/msf-11</span>       
          <ul>
               <!-- data in the same manner -->
            <!-- never stop -->

When each GET to the service its start propagation of the current EVENT window
to the client. And never stop 

the request GET http://newbie.com/feed/msf-11?n=1500 means 
GET 1500 data items from the feed and refuse connection

GET http://newbie.com/feed/msf-11?t=2007-10-17T21:00Z
Refuse connection near the specified time

GET http://newbie.com/feed/msf-11?e=fastinfoset produce the same content using
some kind of binary encoding to compress 

GET http://newbie.com/feed/msf-11?e=plain
NEED to populate the propose schema data per row / number. So this protocol
would be used only by closed, yours proprietary system that feel ok to do this
in such a manner. No interoperability here. 




3)  I don't know what to set up in Content-length, may be we need to work in the
schema PARTIAL GET (that is uses for download manager). 

4) We need to work with KEEP-ALIVE  HTTP 1/1 switch on and use the  Continuation
feature (Grizzly / AcyncWeb / Jetty) 

5) We need to create an StreamRepresenation class, it's simple, and probably use
the write(WritableChannel) methods to produce content

6) We MAY use the direct NIO transfer of repeatable chunks like
                                                 </a>
      </li>
      <li class="feed">
         <a href="http://newbie.com/feed/msf-27";>  

7) AND at the last we may look through the ability to use the restlet with the
specific protocol for this purpose. I suspect that MINA is the  good way to
connect restlet with to do this easyly.

mmm, do anyone try to use 4th item (Continuation) with restlet?



Reply via email to