Scott, they need a River

Scott Karp: “I tried RSS in IE7, and it highlights the true shortcoming of current RSS applications — it’s really not much of an improvement over ‘favorites’ or ‘bookmarks.'”

Bing!

Scott says what I’ve been saying over and over. You need a different kind of aggregator and then you’ll get a benefit from RSS. The way most developers implement RSS, Microsoft included, they are just another form of bookmark, and not much better than visiting the site to see what’s new.

I think this happens because the developers don’t use the RSS apps themselves, they were just told by someone that it’s cool and they should do it, and the net-result is something that isn’t very useful. That, and no one studies prior art anymore. They literally don’t look at software that was shipped before.

That’s why I’m getting ready to ship NewsRiver. To set the bar back where it belongs, where it was when RSS started. To try to get the ball rolling again, in some kind of productive way.

Try it out. Here’s my aggregator. The username/password is snarky/snarky.

PS: I only praised Microsoft for respecting the simplicity of RSS, and I meant it. I didn’t say their software would be useful to anyone. I won’t use it. I won’t recommend it to anyone either. I told them this by groaning out loud much as you are when they demo’d it to me. But what are you going to do? They’re the Biggest of the Big. They know better, and it always takes them three times to get it right. However, even Multiplan 1.0 was a spreadsheet. I’m sure they looked at Visicalc before they designed it. We’ve lost some ground here, I’m afraid.

Since WordPress is open source…

I sent an email to Matt Mullenweg the other day and then I realized I sent it to the wrong place. It should be posted publicly so anyone who knows how to work on WordPress and can write production-level code, high enough quality so it could be included in the main distribution, could see it.

Here’s what I wrote to Matt.

Matt, following up on our conversation on Thurs, I’m interested in developing a strong connection betw the OPML Editor and WordPress.

Here’s how it works — there are lots of elements of a weblog that are editable by outlines. The hierarchy of categories, blogroll, templates, blog posts, the about page, all can be edited as outlines. The blogging tool built into the editor already has these connections, implemented with a very simple back-end. It’s not actually very much work, even though it may seem it is, and in fact the architecture would simplify your code (I’ve written five or six blogging tools over the years).

I’d like to phase out of running a back-end. I can’t do the development in WordPress, but I am happy to provide guidance…

So what do you think?

Amyloo and Frontier’s website framework

A few weeks ago I sent Amyloo a few pointers to the website framework docs for Frontier as a clue as to how to experiment with CSS in the design of a template for the browser-based NewsRiver user interface. Now, she’s reached a point where she’s thinking (apparently) about how to make the results of her exploration available to other users. This is always a tricky spot, if you move too fast you end up with a bad design and either have to live with it, or live with breakage. I almost always decide to live with it, because breakage is too painful, and you also lose the trust of the people you want to work with. So I like to go-slow, think, and consider the alternatives carefully, and most important, look at prior art because there may already be a solution to the problem.

Now Amy is moving quickly (and that’s good) because when I looked at her post a few hours ago I said “Oh she just needs to know about file.readWholeFile,” but now I see she’s found it, and has built a layer around it, but I’m not sure that the layer actually buys you very much. I am also reluctant to create new verbs at the top-level of the scripting namespace, there aren’t many of those, because there always seems to be a context, and this gives you a way to create a new verb to stand alongside the earlier verb, without having to klooge things up too badly. Long story, hard to explain, but top-level verbs are usually not a good idea.

Amy asks if this framework is for Manila, actually it predates Manila. Our software is built in layers, and the layers came in years, like the rings of trees or the brain of a lizard. The stuff Amy is looking at was done in 1996, when Frontier took the turn from being a Mac-only system scripting environment to being a web content management system. Some people still think that was a bad idea, but that was almost ten years ago, a lot has been built since then. Even so, what we did in 1996 was good enough to basically provide the user interface for NewsRiver 0.3, and give Amy a way to explore the wonderful world of Fronteir content management. Adding Manila to the mix would complicate matters considerably, because Manila is substantially more ambitious than the website framework, which is what she’s working in. (Also, Manila is not open source, so that makes it somewhat more difficult to explore this way, unless you have the money to buy a license.)

Now, the question of whether or not the input data should come from the file system or not is an interesting one. There are other choices.

1. It could come from the Internet. In Manila we had a macro called includeHttp (there I go breaking the rule I talked about above, it could have been called manila.includeHttp, so we could have borrowed the concept wholesale wihtout confusion, calling the new macr newsRiver.includeHttp, or some such). If we used something like that here, we could still get stuff from files, because file paths can be represented by URLs.

2. Another possibility is to get the content from the object database. This sounds a bit more hairy than it actually is. You could have a command in a menu somewhere that says Open My CSS, and an outline window would open and it would say NewsRiver CSS. You’d edit it, and refresh your NewsRiver site in the browser, and boom, the change would be instantly visible. Now the user wouldn’t really be sure exactly where the CSS resides. But then they don’t have to worry about where it is. I do this in a bunch of places in the OPML Editor, and so far haven’t heard too many complaints. (Check out the Open Log Outline command in the Community menu for an example.)

Anyway, I’m not sure how much sense this makes. My teaching technique is to just string the words together and let people get confused and maybe read it again and it’ll be less confusing until all this stuff is second nature. It’s very cool that the Class of 2006 is starting to take an interest in the underpinnings. This is a good sign, and it’s a path I’ve been down quite a few times before. I’m hoping that people from earlier incarnations of the community will come back in at some point and express some opinions about how we should evolve here, of course in a constructive and helpful way.

PS: For more clues and philosophy check out the Going Crazy series I wrote for Radio in 2002.