< Previous
Next >

[Comments] (3) Hardy Kingfisher: Keeping in the spirit of "Might as well do a release", I just did a release of NewsBruiser. Yes, you heard right. The last release happened over three years ago, back when there was still hope in the world. People seemed to smile more then... there were concerts in the park... but enough of this sepia-tinged nostalgia. NewsBruiser 2.6.2 is here!

New features include: commenting is less aggravating for commenters (you no longer have to give your email address, a useless feature akin to the way BBS signups used to ask for your home address), comments are less aggravating for administrators (a certain class of HTML-heavy comment spam couldn't be removed via the edit screen, and now it can be), "Today in History" works a lot better for weblogs that have, say, ten years of archives. The del.icio.us export supports the new del.icio.us API. Stuff like that.

As usual, all of the crummy.com weblogs have been automatically upgraded, and since that's a pretty good chunk of NewsBruiser's active installed base, we're doin' pretty well. Of course the crummy.com weblogs have had most of the bug fixes for some time, because all those changes I mentioned, and indeed all the NewsBruiser work I've done for the past three years, were directly motivated by my needs or someone on crummy.com's needs or Sumana's needs.

This is the big problem with NewsBruiser: in the early 2000s I'd add features (RSS 3.0 anyone?) just because I thought they were cool and I liked programming, or because someone had said they might use NewsBruiser in the future if I added such-and-such a feature. The result is a sprawling code base (at least it's modular!), a UI that reaches further than my limited UI design skills can take it, and except for Seth none of those people ended up using NewsBruiser anyway. And now I've got a program that works fine and is very reliable, but which contains core samples from my entire history as a working programmer (which means a lot of ugly code) and has a lot of unnecessary features.

Some of my initial goals (running on Python 1.5.2 because hosting services didn't have 2.x yet) don't make sense anymore. Some (no external dependencies at all) still make sense in the abstract but I think they make NB a lot more complicated than it needs to be. The template system is weird and terrible, and I should be using someone else's template system that has things like loops. I still don't think NewsBruiser should depend on a particular web framework but I would like to keep the data store in a SQLite file. That would make indexing and categories (to pick two examples) much simpler.

It would put an end to NewsBruiser's current infinite extensibility, but you know what. I've used this program for ten years. I'm the person with the strangest needs who actually uses it, and I've got a pretty good idea of what I need from it. So I could just start over, bring in whatever code I can from the old NewsBruiser, and have something shiny and compact that implements the Atom Publishing Protocol.

To the extent there's demand for weblog software written in a reasonable language (ie. not Perl or PHP), NewsBruiser could still be a player. But my hard realization of today is that there's not really such a demand. The vast majority of weblog installations based on free software use Movable Type (Perl) or WordPress (PHP) or LiveJournal (Perl). The people have spoken, the bastards. And life is short, and I've already got an open source project that hundreds of people depend on, which I don't spend enough time working on. So why spend a lot of time improving another project when no one (not even me, really) will appreciate the improvement?

So we bruise on, boats against the current. I'll make the changes I need to make to keep myself and my hostees happy, but NewsBruiser is officially what we call "stable", and has been for three years. I probably won't do any more than small fixes until the Python 1.5 idioms in NewsBruiser start being deprecated.


Comments:

Posted by Aristotle Pagaltzis at Sun Apr 27 2008 12:10

I take exception to the Perl remark, especially at lumping it into one category with PHP. The language is every bit as sensible as Python. (In some ways, more so: scoping is not broken. Mutable closer-over variables anyone? In others, less so. To each their own flaws.)

(Here’s a feature request for NewsBruiser I can’t believe you haven’t gotten before: comment preview. Or at the very, very very least, a note on the comment form about what sort of markup is expected, if indeed any.)

Posted by Leonard at Sun Apr 27 2008 21:16

I really don't want to have a programming language argument, so I'll just say that the deal-breaker for me is Perl's OO design, and hopefully we'll understand each other and it needn't go any further than that.

I've added a list of the supported HTML tags to the bottom of the comment form.

Posted by Aristotle Pagaltzis at Mon Apr 28 2008 13:42

Thanks!

Yes, I understand the OO model complaint. If you care, check out Moose, which is close to the Perl 6 OO model and is emerging as the standard way to do OO in Perl 5. To say anything more about it would be to risk sliding into evangelism, so… just take a look for yourself.


[Main]

Unless otherwise noted, all content licensed by Leonard Richardson
under a Creative Commons License.