[evgen]

What is up with the stacked bar charts used to show performance improvements in various css/html features? I can't really think of a seeing a more useless/opaque chart in quite a while. They make it impossible to do effective 1.3 -> 1.4 comparisons for specific browsers and since no one is running on all five browsers at once the weird style chosen doesn't even coherently communicate an overall point.

[FiReaNG3L]

There's raw numbers at the very bottom of the page

[evgen]

Which you don't see until you scroll all the way down there. What annoys me about the chart is that it is obvious how to do something like this (two bars, before & after, for each browser) and the creators of the page apparently know how to create such a graph because the first graph is exactly in this format -- it is all of the remaining graphs that use the information-free format.

[seasoup]

This is a very standard way libraries have been presenting their speed comparison for quite awhile now. You can look at the individual slivers to get a rough estimate of the size difference, and overall cross browser comparison, and see which browsers are the slowest.

It's a very good visualization, letting one compare the same browser across two version of jquery, an aggregate of browsers across both versions, and the speed of individual browsers against each other on the same version, and individual speeds across the different versions. It's very data packed.

[evgen]

No, it is not a good visualization.

You cannot easily compare the same browser across two versions of jquery because each browser starts at a different point on the Y axis for each column; you cannot easily scan from one column to the next to see the difference but must jump back and forth figuring out start-end points and trying to guess at differences.

It makes it difficult to compare the individual speeds across different versions because the nature of the chart compresses the vertical space to make the larger of the two stacked bars fit into the graph, artificially compressing the Y axis (which measures the only thing we care about in this graph) and diminishing the scale for all of the items being compared.

The only thing it makes easy to see is an aggregate of browser speeds across versions (and since nothing runs on all browsers at the same time this measurement is not very valuable) and as a comparator of individual browsers against each other, which is the second least-important metric in these charts.

[seasoup]

I think you're being deliberately obtuse.

Take the Performance of .css() and .attr() chart. I can tell very easily that in FF 3.5, time to execute cut in half, ie 6/7/8 and Opera 10.10 are ~2/3 what it was, and safari and chrome about 1/2. You can also see that in 1.4 FF comes in third, behind safari and chrome and ahead of opera 10.10, and that ie is the slowest in order or 8,7,6.

It also shows that FF 3.5 has great gains with this release, as it was behind opera in 1.3

That's a lot of information packed into one tiny chart. Yes, diving into the specifics takes some doing, but the primary purpose of the chart is to show the overall gain in 1.4 vs. 1.3. Breakdown by browser is a bonus.

[neilc]

While it is possible to compare the same browser across two different version of jQuery, it's pretty difficult to do so: you need to manually "extract" the slivers from the two different bars and compare them visually, which is difficult and imprecise. I think it would be more useful to have one pair of bars for each browser (old vs. new jQuery version), and then a pair of bars for "total over all browsers" (old vs. new jQuery). That is slightly more verbose, but more useful too.