Radbot progress can continue with everyone working from home, though the face-to-face stuff is good and I will be pleased when we can resume it. (There are only a few things that we need to do other than at home.)
So I may get more time to work on this site!
2020-03-22: I have stopped any featuring of the 'holiday' page for now.
Ads still comprise a huge chunk of the weight of each page on which they are shown, indeed often the vast majority of it. So I have turned them down another notch on desktop pages, reducing the maximum shown on any one page in general, and limiting that still further (nominally to one) for all but the most popular pages.
The aim is to continue to earn some revenue, but with classier ads and less average burden on visitors.
2020-03-22: My attempts to turn down ad load were being largely ignored or overridden by AutoAds, so I have already turned them off even for desktop EOU pages for now. In any case I suspect that what the world needs right now is not more ads.
I have just tried the current Firefox Beta (75.0b2) and it is clearly responding to loading=lazy
. Firefox is much more aggressively lazy, and therefore likely more bandwidth efficient, than Chrome. Firefox 75 beta is apparently not loading lazy images until they just enter the viewport (vs thousands of pixels below it à la Chrome 80). That seems acceptable, though appears to create some jank of its own.
The validator vnu.jar_20.3.16
emerged today as a full release! Hurrah, and thanks to sideshowbarker
!
I have made a couple of very small optimisations which will likely only affect a few edge cases, especially on slower/mobile browsers, but at least should not make anything worse!
First, I have put the src
attribute last in the img
tag so that all the other attributes, especially decoding
, are available early for when the image data is inlined and thus the src
text value is large (many hundreds of bytes). My intention is to slightly improve processing and rendering of such inlined images, and any other images with a tag that crosses a packet/IO boundary.
Second, I tweaked the trailing Share42 social media buttons script at the very end of the body. It was marked async
, and is now defer
in the hope of allowing the DOM to be interactive (etc) a tiny bit sooner. Note: this does seem to push DOM Content Loaded back from lining up with DOM Interactive until the JS is loaded. I will see if anything complains, such as MachMetrics.
I am not expecting miracles from either change!
2020-03-14: I spotted a couple of places where I was still putting something that might be important to layout (in particular a style=width:100%
) after an inlined large src
attribute. Now fixed.