Earth Notes: On Website Technicals (2019-07)Updated 2020-07-19 18:00 GMT.
Having upgraded my Firefox to 68.0.1 I was suddenly experiencing multi-second pauses (maybe garbage collection?) on rendering, entering text, links becoming clickable, etc. So I went for the Help -> Troubleshooting Information -> Refresh Firefox option and whatever crud that was causing the issue seems to have been flushed away. Hurrah!
But also, I must previously have overridden the default text size to be larger than usual, since suddenly I see Firefox's rendering as unhelpfully small as Chrome's, for example.
So I've added to 'desktop'/wide mode (viewport wider than 800px) CSS
line-height:1.5 to see if that helps readability when there is real-estate to spare. The mobile ('lite'/AMP) modes are not touched.
2019-07-21: FFmpeg vs AVconv
At the moment I'm using
avconv, and it is somewhat differently buggy between my older RPi/Raspbian, the new RPi3, and macOS. For some of those
ffmpeg is some sort of alias for
avconv, but not for all.
The arguments continue. But it sounds as if
ffmpeg is currently better maintained, such as from a security point of view, though that's less directly important given that I control the use of the tool and the videos fed through it.
I experimentally switched over to
ffmpeg, but the version on the RPi2 is tool old to be useful, so have switched back for now.
2019-07-20: Search Engine Impressions vs Clicks
There has been a lot of compliant that Google shows snippets on its search pages in response to a user search query, and the user no longer has to click through to the underlying site to get what they wanted.
If you have built a business on getting people to visit your site, from legit news site to crummy scammy Made-For-Adsense (MFA), that may clearly bad news for revenue and viability.
If you have an information site and you get that information to the user's eyes one click quicker, so much the better. And when Google presents it directly in the results, it has some extra credibility.
The poeple complaining somehow feel that Google and the user owe them that click-through inherently (separate from any arguments over copyright in the snippet).
I understand the pain: my peak revenue from Google ads was something like USD2000/month many years ago, now it is maybe £2/month.
But that does not create any right to users' time and effort and lives.
For EOU I see impressions rising hugely while I have been pumping in the schema.org metadata, and adding the AMP site view. These things I think help people see the information that I want to share without having to hit my site directly much of the time.
2019-07-14: HTTPS is Alive!
I have my RPi3B+ on-line and, with the aid of
working on a test Apache domain/site. Hurrah!
2019-07-12: Happy Video
I'm generally happy with how the video support is going, and direct insertion into EOU pages of Gallery video is especially pleasing.
The space saving when transcoding from the original format to
.mp4 can be astonishing. And when cranking down the fidelity even more so. For example, ~78MB original
down to 10MB (
.mp4) down to 2MB lo-fi as below, at time of writing. There is some additional saving when downsizing for a smaller display than the original. A few seconds of video can take a similar size in bytes to a single still as a JPEG. I may yet be able to tune codecs to improve further (ie with slower presets).
It looks like although the
avconv on my RPi2 seems older and more buggy, the version on my Mac may be silently borked instead. The above video was wonderfully broken by the Mac's attempt at transcoding, and was surreal like something from the end of "2001"!
2019-07-05: Better Video support
I have spent longer than is wise already playing with
to create 'responsive' video output better matched to the user's device, eg fewer pixels and lower-fi to minimise data use for 'lite' pages.
Next up is making it possible to use videos directly from the Gallery, copying or transcoding as necessary.
At some point I'd also like to add a short-cut for the repeated re-encoding logical steps, such as full size -> correct width -> lo-fi. Instead I'd pass in the original source alongside the previous element in the chain, and take metadata (eg size in pixels) from that previous item, but recode from the closest-to-source bits each time. That should make for cleaner and lower-bandwidth results.