MIDDLEBURGH

A Technical Post

Thread: previous

After changes last week, a few things weren't quite right ...

1. page took too long to load
2. page not rendered as intended in Firefox
3. displaying previous posts is not intuitive
4. various search engine issues
5. and I broke my Googlemap
    

Page Load Time

Not withstanding the grunt of a visitor's PC and it's connection speed, page load time is ultimately determined by the number of components loaded, their size and where they come from.

I normally check the component load pattern using Ian Holsman's free web performance monitoring tool and based on the chart and table I would look to

1. reduce number of components loaded. This reduces overheads associated with DNS lookup, initial connections etc. A lot of components are loaded by blogger which I can't do anything about. I did remove redundant images and rationalised duplicated function by removing the tagboard. I have decided to use only Haloscan to capture feedback because of its small footprint.

2. reduce size of components: I removed unnecessary code etc in base page,(but not white space); I only display last post since my posts tend to be wordy.I optimised images with picassa (free) and JPEGWIZARD both in terms of dimensions and resolution/color density.

3. relocate components (images) to faster servers/with faster connections: (Under review - will do a seperate post) Initially I stored images on blogger until I realised that they were being optimized on upload with quality degredation. I am experimenting with a couple of other hosts. Load time is dependent on host server speed/capacity and how far down the internet backbone it is from end user. Although data nominally travels at speed of light the further away the source is the more nodes/switching points have to be negotiated ie data routed from a server in Singapore to a user in Japan gets there much faster than same data routed to Finland . This is why commercial companies use Akamai. Ian's tool indicates load problems but times are based on routing to his "location". Keynote (not free) has an equivalent diagnostic tool but with the option to specify target end user location (in my case Dublin). They use a network of robots all over the world. An alternative would be to use a variant on this script and solicit visitor feedback.

4. resquence the component load order: User percieved load time may be shorter than actual: browsers request components in the order that they are listed in base page and render them as available. I push non displaying javascripts etc to end of the page.

Firefox

I forgot to check, and due to some sloppy cleaning up of the CSS, borders where not displaying correctly:I ran the W3 CSS validator and corrected: NB the validator still reports some errors where I have used IE only CSS declarations.

Previous Posts/Navigation

I overlooked fact that when you open a previous post or archive link in Blogger, you refresh the whole page. Normally it seems that only the 'post' is changed but using a tabbed layout, the refresh is more obvious and the user has to reopen the "post" I I am looking at how to get round this but, as a stop gap, have put "help" in the "previous" tab.

Search Engine Issues

My ego isn't hurt by getting low placements in SE results since the target audience for this blog is very select. Nevertheless I do pay attention, if only because the Middleburgh censor is very concerned, and I realised there were a couple of "problems"

Nav Bar Copy in SE results

I didn't expect to see copy from the Blogger Nav Bar in the SE result snippets.I for one would prefer to see my meta description rather than a snippet which starts by asking visitor to "Notify Blogger about objectionable content" etc which will surely put visitors off. A google search returns 112K hits with this problem so it's not just my blog blighted. It is instructive to see how spider emulators such as poodle predictor or Search Whosee the site.

SE Cache Issues

The SE caches for my posts were a mess. I found a variety of really old posts, one which was spidered as the template was actually being changed and a recent post which renders spectularly as a result of using the CSS z-index property in the new template. Since I disapprove of cache bashing anyway I have dissallowed caching and I am waiting for the old copies to dissapear.

Page Not Found

Blogger returns a "not found here page" where

1. visitor opens a SE results link to a page which had been taken down after being indexed (as in this link)
2. visitor opens a SE results link to a republished page where author has change the post TITLE.
3. status of a indexed page is changed from published to draft

With the exception of the example cited I am reinstating taken down pages with MY redirect message !!.

Google Map

Evidently there are issues with Googlemaps and nested/non displayed DIV's and although I found a thread with code, it unfortunately didn't work here so being lazy I fell back to.

Posted at 6/18/2006 04:43:00 pm by David Middleburgh   Permalink

|