The Valid Approach
Back in 1992 or so, when the WWW first came around, the internet was free to roam, and the HTML was kind of messy. This built up over the next few years, and a few years ago (amongst others) Jeffrey Zeldman founded The Web Standards Project. But because there was a lot of 'prior art', that is, jumbled, nonsensical or just plain non-validating HTML, there really is no set benefit of writing validating HTML today, or even use more than FONT tags.
I have a dream of an official Atom (or whatever) Parser. It would be officially maintained for a number of programming languages to most popular systems and setups. Using this, the aggregators could instead focus on features and compete on equal ground.
But wait? Only accept valid feeds? What the f**k, am I out of my mind?
So it may seem.
But. The Parser, the aggregator scene (hopefully), the web log ware scene (hopefully) and the already evolving ease of Atom (or whatever)'s tags would make sure that this would be a goal easily reached. By using a nice, extensible format with easy tag names, this format could really 'have it all'. For hand-rollers, it would be easy to write; for newbies or those who Just Don't Care(tm), sites such as TypePad and BlogSpot would provide automatic, valid Atom (or whatever) feeds, and the latest web log ware would contain a default template or function that would output valid Atom (or whatever) feeds.
The Parser would be centralized to a certain extent and constantly be updated with the parsing rules of the modules produced for the extendable interface Atom (or whatever) will have. Aggregators could easily integrate it and start accepting Atom (or whatever) feeds even when Atom (or whatever) is still in development, and stay up-to-date with the latest changes.
The Dark Side is seductive and tempting
Ofcourse there is a danger that people will start making aggregators that will accept any old Atom (or whatever) feed. However, if we play our cards right here, we could end up with a standard that Just Works, is easy to write, widely implemented and valid by default almost everywhere you look. Then there wouldn't be a market for those aggregators (it wouldn't be a benefit to have a liberal Atom (or whatever) feed parser at any rate) and we would have succeeded.
Comments and suggestions
Comments, suggestions and errata to my flawed vision are welcomed. Be my guest.
[DirkHaun RefactorOk] While I doubt that you can really force it that only valid feeds will be created, I'd suggest that the final specs urge feed providers to validate their feeds and also encourages sites that collect links to feeds to only accept valid feeds.
[wootest] I agree with the validation criteria for link sites. Very nice! The whole idea was to make validation such a natural and ubiquitous part of Atom (or whatever) that it would be the natural - and only, if the aggregators had done their work well, and this is why we need the official Parser I think, to make it easier for them - way to go for people who wanted to create their own feed.
[AsbjornUlsberg, RefactorOk] I like the idea, but I don't think it's doable. Either way, we should at least provide developers of Atom-parsers a neat set of tools to work with, like official an DTD, XML- and Relax NG Schema. We should also provide information on how to parse Atom/XML in different languages on different platforms, and provide URL's to where such parsers can be downloaded. Examplification is very important, especially for the ViewSourceClan.
Also, an official validator (like W3C's Validator) would be extremely helpful.
[wootest] As of now, few but good comments.
Place your name by the option you like best: [OpenPoll]
Yes, follow this plan religiously!
Not everything here is doable, but I'd definitely want to see a lot of it done. AsbjornUlsberg
I'd be happy with just a validator.
No effort should be spent on this.
You're nuts. We should just parse liberally to catch everything.
see also: PaceServiceError