tag:blogger.com,1999:blog-11599526.post2665086965256194782..comments2022-12-15T20:47:08.015-08:00Comments on Looks Good Works Well: Measuring User Experience PerformanceBill Scotthttp://www.blogger.com/profile/12024727845077253669noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-11599526.post-68937924152477436882008-07-02T15:02:00.000-07:002008-07-02T15:02:00.000-07:00It's nice to see such a serious approach being tak...It's nice to see such a serious approach being taken on the question of performance.<BR/><BR/>My company makes an appliance that uses passive traffic capture and analysis to achieve the same goal. This approach allows us to see details that are otherwise hard to discern. For example, we use the first packet of the TCP/IP handshake as our start time. This allows us to measure the time taken by redirects, SSL negotiation, the transfer of the HTTP request, etc.<BR/><BR/>We often find that the performance of our customers' sites can be affected by a number of metrics that occur before the response content is generated by the server:<BR/><BR/>- size and number of cookies in the request headers<BR/>- size of request content (for forms or XML docs)<BR/>- slow redirects leading to the page<BR/>- size and number of cookies sent in the response<BR/>- size of other custom response headers<BR/><BR/>Our approach also allows us to measure performance in cases where client-side scripting is not supported by the client platform (e.g., mobile devices) or by the content itself (e.g., PDF, etc).<BR/><BR/>However, for cases where client-side scripting is supported, we have introduced the ability to analyze "page markers" - i.e., hits generated by the client that carry additional useful data. Our customers use such markers to inform us of milestones that they are interested in capturing - e.g., the "onLoad" event.<BR/><BR/>The case of the "onLoad" event is actually rather interesting. This event marks the time taken by the browser to load the content. However, we have found that this does not always fully describe the time taken by browsers to actually "render" the loaded content. Moreover, not all browsers trigger this event in the same way (as noted in a previous comment by another reader).<BR/><BR/>Client-side instrumentation is still in its infancy (or early childhood, if you will). Perhaps the various browser vendors will adopt (or be forced to adopt) a common API for client-side monitoring and measurement (I'm such an optimist). However, such an effort would have to address the cases where client-side scripting is unsupported by the client itself. With the rise in the use of mobile platforms, this may constitute a significant challenge in the near future.<BR/><BR/><B>Jonathan Ginter</B><BR/>System Architect<BR/><A HREF="www.coradiant.com" REL="nofollow">Coradiant</A>Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-11599526.post-7822068013512934562008-06-24T15:00:00.000-07:002008-06-24T15:00:00.000-07:00You should notice, that Safari fires the onload ev...You should notice, that Safari fires the onload event too early.Anonymousnoreply@blogger.com