53b2867d   
   From: vjs@calcite.rhyolite.com   
      
   In article <14b5a094-6dd7-4e35-b4f0-847bb19e7c43@e15g2000vbe.googlegroups.com>,   
   Jorge   wrote:   
   >Thanks again.   
   >   
   >More details:  my application is a server that receives real time data   
   >from another server, and sends it out to a large number of client   
   >applications.   
   >Data streams in to my app from a single source, and is streamed out to   
   >many recipients.   
   >I am trying to reduce the total bandwidth required by all of the   
   >clients. So, compression reduces the total bandwidth. So, the more   
   >compression   
   >the better.   
      
   Engineering is generally about making choices and numbers.  It is   
   practically certain that "the more compression, the better" is wrong,   
   because the most effective compression schemes would want to compress   
   the entire data set.  There are ways unrelated to compression to reduce   
   the bandwidth needed at any single node in your network.   
      
   I don't know enough about your situation to say for certain, but from   
   what you have said, I would bet at US$0.75 beverage of your choice that   
   typical (e.g. ZIP) compression is a bad idea here.  You might do will   
   with a specialized compression scheme that knows about the special   
   characteristics of your data.  For example, if your data involves stock   
   quotes, some sort of binary code-book would probably work far better   
   than any general purpose compression of ASCII strings of prices and   
   stock symbols.  Or you might pervert something like a Lempel-Zev varient   
   with a static, pre-computed dictionary so that you don't need to start   
   from scratch at the beginning of every chunk.   
      
      
   >One avenue might be compressing the TCP header.  Is this possible?   
      
   Yes, in some circumstances, but almost certainly not in your situation.   
   It is likely that any TCP header compressing that is possible is already   
   being done.   
      
      
   >And yes, I do need a reliable connection.   
      
   Please don't take offense, but the questions you have been asking show   
   that you lack the knowledge and experience to design your own reliable   
   data transport protocol without repeating the mistakes of the last 40   
   years.  The best designers made errors at the time that are now seen   
   as obvious.  Without knowing something about the field, you are doing   
   the equivalent of trying perform brain surgery after talking to some   
   authoritative sounding people at a coffee shop.  You don't even have   
   the tools to decide which of us, if any, are experts only by having   
   stayed in smart hotel last night.   
      
   Make no mistake!  Whatever advise you get from netnews (or what Google   
   calls "google groups") is worth about the nothing (in money) that you pay.   
      
      
   Vernon Schryver    vjs@rhyolite.com   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)