[seam-dev] Seam website optimizations

Christian Bauer christian.bauer at gmail.com
Tue Jun 9 18:38:15 EDT 2009


On Jun 10, 2009, at 00:24 , Dan Allen wrote:

> I have no doubt you have put significant effort into understanding  
> the performance of the website. But there is a new development in  
> the area of performance. Google recently released the Page Speed  
> plugin for Firebug, which I think it's worth pointing at the site to  
> see if there are any new optimizations we can discover. Also, it  
> appears that RichFaces is very guilty in terms of missing the cache  
> and should certainly be discussed at the upcoming meeting in Brno.

We are way beyond running tools to find problems, we know what the  
(fixable) issues are:

Richfaces is bad with a cold browser cache as it's delivering its JS  
and CSS content in dozens of requests instead of bundling the data in  
a few requests. This is obviously going to be on the table. However,  
with a warm browser cache it does conditional GET and cache control  
headers just fine and data is not downloaded again. Simplified: Yes,  
you can optimize the first visit to a page but any subsequent visit is  
not executing any requests and using the cached JS and CSS. The only  
thing that needs to be done is bundling up the resources better. As we  
are not using Richfaces components on sfwk.org on regular pages, this  
doesn't affect the website much.

The Seam resource servlet does not do conditional GET and cache  
control headers. That is bad for people using @WebRemote, as this JS  
is downloaded on every page request, over and over. Also, an  
application like the wiki that uses the Seam resource servlet  
extensively for delivering plugin CSS/JS/IMGs is suffering from that.  
I've mentioned this before on this thread and there is a JIRA issue.  
If someone wants to work on it, contact me so we can coordinate. This  
has implications for REST integration as well.

(Yes, Page Speed is great because it analyses some very obscure  
possible improvements. However, none of them would have any real  
impact compared to the two issues above.)




More information about the seam-dev mailing list