We’ve been funding work on bloged, originally a Gosling weekend hack, and needed to checkpoint how it’s doing; so I went and tried out a bunch of blogging clients, with very mixed results. So far I’ve tested basic writing-in-the-browser, bloged, Ecto, MarsEdit, and w.bloggar. [Update: I’d misconfigured Ecto.] [Update again: I changed my mind and totally rewrote the conclusion; ain’t this medium wonderful?]
I’ll give the conclusion first, then the methodology, then the notes on the individual tools.
Conclusion · My initial reaction was pretty horrified, along the lines of “and people put up with this?” But then I realized that I’m severely spoiled. I spend my writing time either in Emacs or a word processor (formerly MS Word, increasingly OpenOffice), both fantastically sophisticated environments, the result of decades elapsed and person-centuries invested. The blogging tools are nowhere near as good. But the best of the bunch—in my experience, Ecto and MarsEdit—are good enough to get work done.
So here’s the situation: the vast majority of bloggers are stuck writing into a browser input field: which pretty well sucks; they’re writing with one hand tied behind their backs. A small minority are using tools, just now getting good enough for daily use, which are the product of cottage industries. At the same time, the infrastructure, notably in the shape of the Atom protocol due later this year, is really falling into place, so that you’ll reasonably be able to expect to develop a tool that’ll work pretty well with any publishing system.
In case it’s not obvious, I smell opportunity here; major opportunity.
Methodology · I’ve got a little workbench over on blogs.sun.com, which is based on Roller and supports the MetaWebLog API. For each authoring tool, I fetch the existing posts and see if they can be modified, then I create a new post with at least one link and at least one picture.
Is this enough? I think so. There are a million bells-and-whistles you could ask for, but I’m a firm believer in 80/20 points, and if a tool makes the easy things easy and gets out of the way, then I bet it’ll catch on and grow the extra niceties.
Also, while I’m reasonably good at figuring out software, I’m impatient and if I try two or three obvious-seeming ways to do X and they don’t work, my conclusion is “You can’t do X” which is I think compatible with the way most people approach new software.
I’m sure there are lots of other candidates out there and I’d be happy to give them a try, let me know and I will.
Editing in the Browser · In the ideal world this would be the ideal solution, I think. Blog-reading is already in the browser, and it seems obvious that feed-reading will end up there too, one way or another. So why not blog-writing?
Unfortunately, at the moment, writing stuff in the browser mostly sucks. It’s primitive, it’s clumsy, and (mostly) doesn’t auto-save, so it’s real easy to lose your work. Seriously, if you’re writing something big and serious that you care about, consider doing it offline in your favorite word processor or text-editor, saving often, then using cut-and-paste to transfer it into the browser.
Sucky HTML · If you do a view-source on my workbench, doing all this cross-editing with multiple clients introduces some serious HTML rot, it’s a wonder that the browsers actually manage to display what comes out.
bloged · Henry Story, who’s doing the engineering on this, has done some radical back-end redesign, after deciding that the right way to persist weblog state was in an RDF triple store. I wasn’t convinced, but it seems to work. He’s wired in Kafenio to do the HTML editing, and I hit a couple of buglets but it seems to work OK.
At the moment, it can’t manage to post pictures for some reason. Also the Kafenio dialogs are kind of opaque and have weird asterisk (‘*’) characters all over the place.
I don’t know of any other plausible clients that are Java and hence portable, so I think bloged is going to have legs, and is worth us funding some further development. Particularly when the Atom Publishing Protocol becomes ubiquitous.
Ecto · This is a client, compiled separately for the Mac and Windows; I tried out the Mac version. It has a pretty nice interface, but didn’t play nice with the blogs.sun.com Roller deployment, until Dave Johnson guessed that I had somehow misconfigured it for Movable Type instead of MetaWeblog. Don’t know how that happened, operator error I guess.
Anyhow, with that settled, Ecto seems to do pretty well the right thing; nobody would call the interface beautiful but it’s lightweight and snappy and gets out of the way. Drag-&-drop pretty well just works. I can see a lot of ordinary people spending their authoring lives quite happily in Ecto.
MarsEdit · This is another Brent Simmons production, and as such, Mac OS X only. I’m well-known to be one of his fans and a heavy user of his software.
MarsEdit had by far the slickest packaging and configuration and so on of any of the packages I looked at; it had no trouble posting, linking, editing previous posts, and so on.
Some things are a little less than intuitive; when I want to insert an image or a link, it isn’t obvious to me that I should pull down a menu labeled “HTML Tags”; mind you, the menu is clean and logical and works well, once you discover it.
I think the editing environment will be a little raw for some people; it has no pretensions whatsoever to WYSIWYG-ness, so if looking at naked HTML tags bothers you, MarsEdit is not for you. Since I normally do my XML editing in Emacs, this doesn’t bother me at all.
w.bloggar · What a weird name for a piece of software; but it seems to stick to the brain. w.bloggar is Windows-only and has an OK-enough user interface; it was the only one where I found a preset for Roller servers.
The editing interface was pretty spartan, the insert-picture dialogue was completely baffling, and when I hit “upload file” it popped up a dialogue asking about FTP settings.
Once again, I suspect that Marcelo and DaveJ could work this out, and that doing this might be a good idea.