Re: [wsrp] [wsrp-wsia] [change request #249] Class name length impacton performance

Title: 

I agree with Rich on the clarity thing. However, I wouldn't have a
problem with shortening the prefix from 'portlet-' to 'p-' and keeping
the rest of the class name the same.

Mike, it's kind of funny that the document size appears to be an issue
when you are considering encoding the navigational state of all the
portlets in the page into every URL in the page. Just kidding ;)

Alejandro

Rich Thompson wrote:

 

  To keep this reasonably in
perspective, files with CSS definitions make wonderful items for
caching and so the download time impact is lessened significantly.
Also, the length of the class names is a minor part of the length of
the CSS file. 

  

  The class names in the content is
unlikely to add over a k in overall file size and therefore be a minor
portion of the performance question. 

  

  Net: Class names should be as short
as is reasonable as they do impact End-User performance, but clarity of
what the class is for should not be sacrificed in order to shorten the
name. 

  

Rich Thompson 

  

  

  
    
      

         

        
        Rich
Thompson/Watson/IBM@IBMUS
        
03/26/2003 10:09 AM 

        
                 

                To:      
  

                cc:         

                Subject:        [wsrp]
[wsrp-wsia] [change request #249] Class name length impact on
performance
      

    
  
  

  

  

  

This came from an internal review by our "performance police":

  

Document: Spec 0.92 

Section: 10.5 CSS Style defintions 

Page/Line: 72/15 

Requested by: Michael Freedman 

  

Old Text: 

portlet-*

New Text:  

shorter class names

  

Reasoning: 

Please strongly consider using *much* shorter names for these CSS
elements.  We already have network latency/transmission issues due to
the size of our existing CSS files because they cannot utilize browser
compression. 

   

Long names in the CSS definitions lead to large CSS files and noticably
slower download times.  Additionally the "source" HTML pages that refer
to the CSS's will be noticably larger as a result of these long names.

  

To give a concrete example, consider a centralized Portal server
configuration, ie HQ/MOC, with a remote client in Singapore, operating
over a fairly standard 200ms latency network.  Transmitting a 10k CSS
file will cause ~6 round trips due to network startup costs. 

   

Overall loss due to network Round trips = 220 * 6 or ~1.3s