ToddG wrote:

I think blog software is hard; there's a kajillion half-way done crappy blogs. People start them with great dreams and hopes, and then they lose attention and they never get finished.

But, I think there's a few people twiddling about with the idea, maybe in and out of Pylons itself, but with pieces that are WSGIish. If it's broken up into pieces, it's easier to imagine finishing a piece instead of having a half-done chunk, and for other people to pick up where you left off.

Here's how I imagine it could work out:

* Tagging/indexing piece. This piece is notified when resources appear or disappear from the site. This gets metadata about the items to create the index aka archive. If it also gets tag information, then it does categories or somesuch. Maybe it would be sufficient to read atom feeds. Maybe the front page would be generated by this, using some query (10 most recent items).

* The atom piece. This gets items from somewhere, probably using the Atom Publishing Protocol. Maybe the tagger watches this, maybe it knows how to ping the tagger when items appear.

* The publishing piece. This is the admin side. Maybe it needs no persistence, maybe the atom store can do everything. One thing I'm unclear about with an atom store is queryability. Obviously you need access to more than just the 10 most recent items when you are administering the blog. But it can't very well serve up a complete dump of the site, even internally. Maybe the atom store would need to be extended with some querying interface.

Anyway, I think it can mostly be a web frontend to the atom store.

* Authentication of some sort, in front of the admin side. You can be lazy and start with a single account, maybe just configured, and make it fancier later. The admin side doesn't need to know much about how it is protected.

Inter-component authentication needs to happen in some fashion, but I guess that's separate.

* Comments. I don't really know how this interacts with an atom store; I don't think there's any obvious way to connect things together this way (at least nothing occurred to me when I first read the atom spec; I haven't paid much attention since then).

Easier just to do it as a simple piece of middleware, I think, and not worry about integrating it too closely with the other parts of the system.

* Maybe a CAPTCHA or antispam service of some kind.

* Full text indexing service, similar to the tagging but with a different storage.

* Maybe some kind of templating. Preferably something that doesn't expose the full complexity of the various underlying templating languages. Similarly something to manage CSS files and whatnot.

* Something for images, but I'm not sure what. They aren't embeddable into Atom, so there needs to be another mechanism for them. More generally, any media.

... I think that's all the pieces that would be necessary. Then of course some glue app that pulls them all together into a cohesive whole. Doesn't seem terribly hard, I guess, and it covers the important features.