I noticed a lot of odd `/rss.xml' requests in my 404 log. As it turns out,
it was my fault. This both fixes it and adds a redirect in case someone
tries to do this manually. I suppose that'd be convenient.
Wow. I had wanted to spend less than an hour on a response, and instead I
wound up writing my largest article since the NSA revelations and
GHS. Hopefully others find this useful.
I've been sitting on this for weeks because I didn't have the time to finish
final proofreading and changes. I need to release this before I sit on it
for another couple; I have to start working on my LP2019 talk soon.
* post/2019-02-18-ghcq-exceptional-access-e2ee-decentralization-reproducible.md:
New post.
* src/papers.rec: Add post to top.
I'm still debating whether to include the full text of the post within the
RSS feed, since some of them may be substantial (like the one I'll be
posting soon that I've been sitting on for a couple weeks because I'm too
busy/lazy to do final editing).
* src/rss.sh: Add "(Read full post)" link.
The inlined CSS intended to make the stylesheet applicaton less jarring on
slow connections was placed _after_ `style.css', which was causing it to
take precedence over the mobile layout. Silly mistake, and not good. And
it went unnoticed for too long; I didn't visit my own website on mobile for
a bit.
Sorry, mobile people!
* src/header.tpl.htm (head): Move style.css link below inline style.
The idea here is to provide as little CSS as is sensible for the initial
page load to be styled in a layout similar to the final layout. This
initial styling may be briefly visible on a slow conection.
Slow connections can happen for a variety of reasons. For example, I'm a
Tor user, and connection speeds vary. Mobile connection speeds can also
vary wildly.
This adds a few hundred bytes, but I was able to cut it down quite a bit,
and I don't find this to be unreasonable relative to the other data on
each page.
This automates creation of the header and footer. Previously I modified
them manually and they got out-of-sync.
This is deployed to a different location on my webserver, even though the
public route is `/projects'.
Rather than displaying the hash separately, this just makes the date a link
to the source code. Until I display a modification date, this will also
make it easy to see the history of the file.
This website honors the user's default font settings (both to be kind and
for accessibility reasons). Consequently, the responsive layout is based on
character units (ch) rather than pixels.
We have two sitautions to account for:
1. Old posts had both uppercase and lowercase letters in slugs; and
2. Some ids changed.
Lighttpd can't convert to lowercase and having a bunch of separate redirects
in my webserver configuration for the id changes is messy. So, this script
is intended to be called only when a post contains an uppercase character in
the path.
I had wanted to avoid _any_ sort of dynamic scripts. Oh well.
All other redirects are handled in the websevrer configuration (which isn't
part of this repo atm).
Rather than having Pandoc generate the id, which has the potential to change
over time and cause 404s, let's just generate the slug from the filename so
that the ids will never change. This also solves the awkward question of
what the filename should be, since it was previously something arbitrary.
This mass rename was accomplished via this simple shell script:
for p in *.meta; do
slug=$( recsel -P slug "$p" | xargs basename )
mv -v "${p/.meta/.md}" "${p:0:10}-$slug.md"
done
with minor manual tweaks where I saw fit. Of course, now I have some pretty
long filenames, which is undesirable.
The next step is to compare it with the slugs currently on mikegerwitz.com
and make them match. That's the next commit, and should be pretty simple.
I didn't originally intend for all of this to be in a single commit. But
here we are. I don't have the time to split these up more cleanly; this
project is taking more time than I originally hoped that it would.
This is a new static site generator. More information to follow in the
near future (hopefully in the form of an article), but repo2html is now
removed. See code comments for additional information; I tried to make it
suitable as a learning resource for others. It is essentially a set of
shell scripts with a fairly robust build for incremental generation.
The site has changed drastically, reflecting that its purpose has changed
over the years: it is now intended for publishing quality works (or at least
I hope), not just a braindump.
This retains most of the text of the original pages verbatim, with the
exception of the About page. Other pages may have their text modified in
commits that follow.
Enhancements to follow in future commits.
Mike Gerwitz