we concept I will likely exclusively facilitate outside the yard?

You are right that the effect is *small*. This sort of optimization isn’t something a small site needs to run out and do. However, a larger site and a larger dev team can certainly work this into their long term plan and will make an appreciable, if small, effect on the load time which will be magnified by the amount of traffic.

In practice, browser can spend about 50% of time (or even more) sending requests and waiting. I’m preparing next blog post with illustrations what’s happening during page load.

Impact of embedding stylesheet is another story for another blog post.
Browsers have not so large disk cache by default (Opera 9.60 has 20MB – probably not covering even daily traffic).
Analyzing http logs so far I have seen that usually cache objects live in cache no more than 1-2 days.

The main delay would still be from sending the css content over the wire – which you’re doing either way if you embed the file. The only thing you save is transmitting a couple hundred bytes that would have been used on the HTTP headers for the extra request, the time of which could be potentially as much as the overhead of calling into PHP to load the stylesheet.

Not to mention that embedding the stylesheet would cause longer load times for browsers that do show content as soon as it’s received (Opera).

Do you have any actual benchmarks on this having a measurable effect that would justify the work?

Unfortunately, both IE and FF block rendering till all CSS files are loaded. So the problem is still actual.

Anyway, isn’t this all made redundant by the fact that HTTP and most browsers support keep-alive? That already removes the overhead of creating new TCP-IP connections for every resource on the same server.

Seeing that, I don’t understand the benefit of inlining the stylesheet on first load. Not only you’re mixing in javascript and cookies, which both can be turned off, and you’re making assumptions about my caching preferences, but you’re actually downloading the stylesheet twice (once inlined, then again postloaded) – actually using twice the bandwidth, just to save a couple of bytes on the HTTP headers. What’s the point?