無題遝事

Mojeek is cute in many ways. Sadly, it’s not for anything that isn’t English. It fails even more badly than DuckDuckGo used to back when I started using it.

After a cursory review of what choices of general purpose web search are there today, I would still recommend DDG. It was always more about it hopefully not colluding with Google, not about any other claims it made. The only problem I have with it is lack of ability to search for fixed phrases (quoting does nothing for Duck) and I ended up reluctantly supplementing it with Qwant from time to time.

I opted in for the new layout on Wikipedia, and, surprisingly!, I don’t hate it.

Anki is such an awful bloatware that Debian maintainer have been unable to figure how to build it for over 2 years and it’s unlikely to change any soon. But it’s not without a reason everyone recommends it. Couldn’t we have something striking a better balance between feature-fullness and complexity? Everything was better on 640K.

寝ようかと 思っていたら 朝が来た

Thanks to Google’s change in policy, I might be harder to reach by the usual email from next month. Please use my home address (mmax) for things that cannot wait a week or two.

As absurd as it is to consider an open-source MUA a third party, I thought it won’t be that big deal as OAUTHBEARER method was standardised in the meantime and mutt even ships with a script to use it. Sadly, Google’s refresh tokens live only 7 days unless you are an organization or actually acting as a third party to other users and pass some verification of your ‘app’. This makes it extremely unpractical; I could just use the webmail interface as this point if I have to constantly re-authorize it through a web browser. Especially since the authorization screen is kind of broken.

それは禁じられているよ、デュオ。Line breaking is hard in any language.

Generally browsers can’t always do the right thing themselves even if the language is correctly tagged (it isn’t on Duo), but nowadays they support line-break:strict which is a good start and it would help with the above.

ハトとカササギを組み合わせるとカケスになるだろう

Lighttpd has this nice option that allows providing the Content-Type header for static files in their extended attributes instead of guessing by the extension in the name. I rely on this feature for paths without extension like mwgamera.name/blank to work. At some point a configuration option was added to allow changing attribute name to take it from, and also some other changes happened that I am too lazy to check beyond noting that extended attributes are mentioned in the changelog.

From the documentation:

mimetype.xattr-name = "Content-Type"   (default)

mimetype.xattr-name = "user.mime_type" (freedesktop.org convention)

Can you spot the problem? Hint: among ‘some other changes’ listed in the changelog there is one to prefer getxattr over attr_get. It’s not surprising that it suddenly stopped to work with the upgrade to debian 11 bullseye. The version of lighttpd shipped in it does indeed default to "Content-Type" without a namespace prefix which has no chance of ever returning anything but error when using getxattr. Explicitly setting mimetype.xattr-name = "user.Content-Type" restores the previous behaviour. (Though one would probably want to use the freedesktop.org convention in new deployments.)

mwgamera:

Wikipedia still/again hates HURC

crabs are cool