gzip compression

3 comments leave your own, follow the feed, or trackback

1. Alex Leigh says:

   January 5, 2004 at 10:16 pm

   The irony is that the target audience for a page compression strategy like this is analog modem users, and their links are already compressed using protocols such as MNP5 or (hopefully) V.44. gziping the text will actually add overhead to the this sort of link because gzip is unlikely to do a better job than V.44

   <p>There is an easy exercise to prove this; go start a large HTTP download for a binary file and then for a large text file. The CPS transfer rate will be higher for the text file. If not then the modem is not negotiating properly or is setup poorly.</p>
   
   <p>Cable modems are (usually) uncompressed transport links, and so there is a slight benefit in providing the compression; but of course someone on a cable modem can load the pages so quickly that this is only a benefit for  extremly, unusually large pages.</p>
   
   <p>Then there is the CPU overhead; gzip isn&#8217;t the most expensive compression algorithm ever developed but it will add latency to the connection, which may slightly delay the browsers ability to interprete the page; particularly browsers that do not perform decompression on the fly but rather wait for the entire file to be downloaded before rendering it.</p>
   
   <p>But while we are concerned about those poor, poor souls who use dial up, let&#8217;s not forget MTU; decreasing your MTU on the dial-up system to 296 is more appropriate for modem use and even though it actually decreases overall throughput for block transfers, it makes the link feel more snappy and reduces latency.</p>
   
   <p>aleigh at tessier dot com</p>
   
2. lee says:

   January 5, 2004 at 10:38 pm

   oh, the irony!
3. MidGunner says:

   January 9, 2004 at 9:44 pm

   so is my page loading faster or do i just think it is loading faster?