ietf-openproxy
[Top] [All Lists]

RE: OPES Status Update, Drafts, ICAP

2002-09-25 00:10:44

Hi Arumugam,

2. One concern over the request modification.

   Icap client sends  the client request to icap server. The 
icap server
   could sent an adapted request , an HTTP error message
   indicating the client request is prohibited or ICAP error msg.

    The icap server sends ICAP error msg due to an error in ICAP
    semantics such as ICAP request is not correct or service not
available.
    In this case , the icap client can either forward the original
    request without any modification to OS or send an error msg to 
    client stating that the service could not be provided.

    It would be useful if the draft says how to handle this scenario.


There is no clear way how to handle this.
Examples:
 - If a virus filtering RESPMOD service sends an ICAP error, the original 
request is probably better not bypassed
- If an advertising removal RESPMOD service sends an ICAP error, bypassing the 
original response is probably the prefered option.
- If an Internet Access Management REQMOD service sends an ICAP error, there 
will be companies that would allow a bypass while others prefer to block the 
request.

So usually the bypass flag is an option to set with the ICAP service that is 
defined at the ICAP client. It does not belong to the protocol specs, I think.


3. The draft is not posing any prerequisite to the tools used 
to provide
   the actual services. Based on the protocol , it seems to 
be valid to
   have few prerequisite on the service tools.

   1. The tool has to be run in non-interactive mode. GUI interface is
      not needed.

   2. The tool should be started by cli. Allows to pass command line
      arguments.

   3. In the simplest form, each instance of the tool serves only one
      request.Considering the overhead due to the size of the 
tool and 
      time to load and make it ready , it would be fine to 
make a single
      instance of the tool to serve multiple requests either
      sequentially or parallely. I am not sure how far this is
achievable.


Does this belong to the protocol specs?


4. It could be possible to make use of the ICAP protocol for real
   bandwidth savings.This is possible by providing a service 
   called decompression.

   For example, consider a scenario.
   The client makes a request to html page in OS. Due to ICAP 
,a request 
   with  a url which points to the compressed copy of the html page is
   sent to OS. This is done in REQMOD mode.

   OS sends back the compressed data. The compressed data is sent to
   ICAP server which can decompress and send the data in the format 
   requested by the end-user. This is done in RESPMOD mode.

   Thus the bandwidth needed to transfer the html page is 
reduced to the
   level of compression.

   If this service is considered as a valid one, then this 
service could
   be mentioned in the draft.



While having good examples in a draft is nice for easier understanding, it is 
impossible to provide the full list of all valid and usefull services.
I think a place like the ICAP forum is better to collect all ideas for ICAP 
services.

Regards
Martin

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