This was a humorous and informative quick talk
WebPage Test http://velocity.webpagetest.org/ – More features: Ability to filter out certain content e.g. test without BV or other third party, download dynatrace data from their Virginia location, tcpdump to import into wireshark etc.
I’m at Velocity http://velocityconf.com/velocity2011
PPK is the owner of the invaluable QuirksMode.org and has been digging into mobile browsers for the last few years – This research produces some quite disappointing results with layers of complexity in browsers that you would expect to have some consistency – “There is no webkit for mobile”
Anyway, he gave a talk recently and has published the slides at http://www.quirksmode.org/blog/archives/2011/03/gomez_webinar_s.html
Some of the data was eye opening for me.
I have had the conversation several times recently about jQuery, dojo, YUI etc.
I’ve come to the conclusion that the main differences are that things like dojo and YUI are frameworks (I differentiate by them having a packaging system) whereas jQuery, prototype etc. are libraries (Although jQ will roll the UI library).
This blog post summarizes it nicely. jQuery is for n00bs
I do love jQuery for quickly getting things done on a website and love to read their code and bugs to collect the oddities of various browsers, but for a packaged componentized system dojo wins for me.
Some fun with the HTML5 history API
The link to the blog post includes descriptions, video, slideshare and a summary of how the Alexa Top 1000 stack up. There is also some information on how Facebook tackled some of their performance issues in this area.
While I like the overall approach in the article and have looked with interest at Nicole’s (aka, “stubbornella”) work, I am also hopeful that the adoption of systems such as Sass or LESS will also take this forward.
Ajaxian posted this one and the attached document is worth a read. They have a good summary and if you read the document there is a recommended practice, but you should be aware of the shortcomings of other methods.
I’ll look, in the future, at how this affects the list of sites we are reviewing.
Download the article from Stanford Web Security
Now we are getting down to the nitty gritty of e-commerce websites. Once you are through the shop path you are at the point where you can really start to purchase things – The product page. Because there may be a variety of information about the product, zoomable views, additional images, personalization options and shopping functionality – These pages tend to have more scripts and so tend to be larger than other pages.
Now, we may not always be comparing like with like here. product pages on a site may have different capabilities, some may have video, some will have customer reviews (More on the 3rd parties providing these services later), some may have multiple product options and we have already mentioned alternate / zoomable views.
Items of note wile doing this post were:
- Macys site was unavailable when I first tried it
- PageSpeed does not like some pages e.g. at Crate & Barrel and Best Buy and so i could not get a PageSpeed score
Items of note in performance were
- GAP – Similar number of page items to Amazon and a faster return visit time – Although, I am mainly looking at the Document complete point here. If we wait until document loaded they have had quite a bit more time and page weight, but still a good time to DOM complete and good postponement of processing. We’ll look at that in detail another time
- Sephora again comes in with a high number of requests to get to DOM complete for subsequent requests
- A good few pages took more than 10 seconds to get to DOM complete on our ADSL from Virginia test
- Williams-Sonoma got a rather easy page with no video or other heavy pieces. Will re-run on a video page, however the cores on Y!Slow and PageSpeed are still good
In this post I have collected the performance of the first category and then a further sub-category from the sites I am looking at.
I let webpagetest test run each test 10 times to get a better average – This is still not that scientific but hopefully will be a little more representative. At some point I will go back and do the same for the home pages.
Now we are off the home page, we should be getting into areas that more represent the architects and engineers work.
Some things to call out on the Category Pages are:
- CB2 has the largest initial payload at ovr 1MB. The main component of this appears to bet the VisualCart.swf coming in at 493K
- Crate & Barrel and Sephora come in with over 90 repeated requests when the category page is called for the 2nd time. This is nearly 3 times more than Amazon, which at 33 still has nearly 20 more requests than the next highest.
- Amazon and Sephora both have over 100K in requests for a subsequent download. This seems a little high, although it will take a little more research to identify what is being reloaded that is not being cached
The spreadsheet is embedded below.