I welcomed Google's termination of its Reader service in July 2013 as an opportunity to curb my addiction to online news. Unsurprisingly, after a brief period of improved productivity, I found myself in search of a new RSS reader. These were the alternatives:
Cloud-based services . Several providers have jumped in to fill the void left by Google, but I did not want to deal with end-of-support issues ever again. If Google cannot provide a long-lived feed aggregation service for free, then who can ?
Browser add-ons . Since I use completely different browsers on many devices, synchronization would have been a major problem.
Self-hosted server . There are plenty of open-source RSS reader projects with a web interface, but it looks like they all require Java/PHP/Perl and a full-featured transactional database, a.k.a. bloatware and dependency hell.
Eventually I wrote a quick-and-dirty, command-line-friendly feed parser which addresses 80 % of my requirements with only 100 lines of shell script and XSL rules.
A generic UNIX/Linux server
xsltproc (from libxslt). Regular expressions would work reasonably well, but now that XML has infiltrated everything, xsltproc comes for free, so why not use it ?
wget.
Some way to export HTML output to multiple devices: a self-hosted web server (preferred), or a web hosting account, or simply sendmail.
A HTML viewer.
Feed subscriptions are stored in a plain text table (example). Note the emacs Org-mode-friendly format.
The last column specifies how each feed should be displayed:
- rawhtml: Copy HTML from feed entries (dangerous).
- noscript: Mangle <script> tags.
- notags: Hide all HTML tags.
- (default): Display the title only.
rss2table.xsl and xsltproc transform XML-formatted RSS and Atom feeds into plain text with one item per line.
rsskiss_poll.sh reads the subscriptions file, retrieves feeds, and formats unread items into a HTML document on standard output.
(optional) rsskiss_poll.cgi is a simple wrapper for rsskiss_poll.sh. It can run either as a cron job or as a CGI script to refresh feeds manually.
(optional) rsskiss_mark_all_read.cgi lets me mark articles as read from within a web browser. Alternatively I could trigger it with an email reply, or with cron, or on the command-line.
wget http://www.pabr.org/rsskiss/rsskiss-1.0/rsskiss_poll.sh wget http://www.pabr.org/rsskiss/rsskiss-1.0/rss2table.xsl wget http://www.pabr.org/rsskiss/rsskiss-1.0/test.org sh ./rsskiss_poll.sh test.org > new.html firefox new.html
rss2table.xsl has been tested with a few dozen feeds from various content-management systems, but it is not guaranteed to work with every variant of RSS and Atom.
Displaying third-party HTML code from a trusted HTTP server may be a security concern, depending on how the server and the browser are configured to handle security domains. rsskiss does try to sanitize untrusted data, but it probably does not catch every HTML escaping trick.