On the other hand, with files visible for enduser, less files is nearly always better. Thus, plugins using CSS- and js-files should a) optimize those files and b) make it easy to take control of those files for combining and minification.

Not related to optimal code, but optimization otherwise – Joost's sociable is a good example, since it's one of the few “social bookmarking” plugins that uses CSS sprites, making it much better for speed than many alternatives. (and it also has an option to not load the default CSS for ultimate control)

* Torbert (All In One) and de Valk (e.g., Sociable) tend to code in very few, very long files.
* Godley (Redirection) tends to modularize into many dozens of files.

Which is more optimal?

It's completely possible to have a blog load really fast even with images, taken that the images are optimized and not too large.

Also, in reality, loading times with pages that have images in them are often better than the tools show, because when you have the width and height defined for the images (or they are in background and are loaded via CSS), loading of the images does not block page loading = the page shows to the reader quickly and images are loaded afterwards.

But I like that WebTrends tool! It's new one for me, thanks for sharing! The option to choose a city is AWESOME.

I've been personally using http://tools.pingdom.com/ and http://www.webpagetest.org/ myself (in addition to Firefox plugins like YSlow and Google's Page Speed)

Pingdom tends to show relatively realistic results (often the “worst case”) and with WebPageTest you can choose couple of locations as well and get cool breakdown charts and suggestions what to optimize.

The tool at WebTrends is really awesome too – http://www.uptrends.com/aspx/free-html-site-pag…

You can select from about 30 different geo-located servers. So for we Southern Hemisherians, you can select a server closer to home (Sydney or Singapore) to getting a more realistic representation for local visitors.

NB: I don't work for them! I do love their free tool though.

The fact is that 99.9% of the plugins are not using the most optimal code. There's a reason for that: the plugin authors want to keep their code readable and editable, and for most, they just don't know better. With thousands of plugin authors out there, most of them are not that good at coding.

And with good I mean those who do what they want using the least amount of code and resources. When it works and does what the plugin author had in mind, they publish and leave it be – there's some seriously sloppy code in some popular plugins, let alone themes!

So hard data or not, people should pay attention to the number of plugins they use, especially if the plugin adds something to the sidebar (widgets), posts or pages. In addition to that, some optimization needs to be done by hand, or on the fly.

But my opinion is most people don't know what they're talking about when they say “plugins are slow.” They don't even know they don't know what they're talking about. They're just echoing what they have read elsewhere. My opinion is pretty strong on this point because it's pretty clear most of the people saying this don't have the technical expertise to evaluate their claims.

In any case, someone needs to run some performance tests on a suite of “standard” plugins and collect a lot of data. A lot of data, real numbers. It won't be me this month, don't have the cycles for it.

In the meantime, I tend to take a pretty close look at the source code for plugins I suspect might be slowing me down. I'll also use YSlow to check. I have found that YSlow doesn't always find repeatable behavior. That is, sometimes different plugins run at different speeds. Weird. Worth checking into. And I hope you do!