Ian Bicking: the old part of his blog

Standalone WebKit

(Were you looking for Javascript Convert?)

I checked a refactoring of Webware into the Webware Sandbox. Its goal is to make Webware easier to install, configure, and administer, as well as making Webware applications easier to distribute. You can browse it online, or view the README.

It's been sitting on my hard disk for a while now, and I thought I'd put it out there even though it's quite rough. I think this addresses some of the biggest problems with Webware right now, and giving it the potential to be more accessible.

Created 05 Nov '03
Modified 25 Jan '05

Comments:

this could be the last chance for me to use webware, i don't wanna sound agressive or anything, i just want to share some thoughts about ww.

i've spent months playing around with ww but recently i've started to hate it, since i need more freedom in the web app layout (i'm considering to have the sitemap in a separate conf file).

what i find frustrating in webware is that it caches everything. yes, i known, turn on "autoreload". :)

...and another irritaning thing: one-servlet-per-file. i just find the mod_python publish approach a lot more elegant.

if this version is more flexibile than the original it would be cool to give it a try.

i'll keep an eye on it, i promise.


# dee

Webware already has a lot more flexibility than people often appreciate. Webware's Servlets, and the Page class that people usually use, are just the beginning of building an application -- application and site customizations are an important part of adapting it to your needs.

This could probably stand to be better documented, but it's difficult to say how -- it's more like presenting Webware recipes than specific documentation you can follow along with. There's some stuff along this lines that people have contributed to the Webware Wiki.
# Ian Bicking

Not to belittle the work put into Webware since the early days, but Webware seemed to be ripe for a refactoring. As for whether mod_python is more suitable for certain kinds of Web applications, I'd say that Webware provides enough interesting functionality for you to build a plug-in for other kinds of resources that present themselves in different ways.

However, mod_python does seem to step more out of the way in certain areas, whereas Webware seems to make certain decisions for you (eg. resources must be "concrete" files) whether that is what you want or not.
# Paul Boddie

Webware has some changes to URL handling in CVS (and in WebKit standalone) that are quite flexible, which would allow you to introduce considerably different ways of representing a servlet. (Servlets are still core, but the actual servlet interface is so minimal that it can serve as the basis for pretty much whatever you want to do)

# Ian Bicking

Ian,

1. I have tried to trackback to your stories a few times but it does not seem to be working. Also typepad's quickpost bookmarklet does not pickup your trackback url automatically (I guess possibly/probably related).

2. Although I don't yet use webware I have found this exciting progress, but there does not seem to have been a great deal of reaction on the mailing lists. Can you see this going forward to become the standard webware?

Dee,

"...and another irritaning thing: one-servlet-per-file. i just find the mod_python publish approach a lot more elegant."

In large projects with rapid change and a sizable team I find this is a big advantage.
# DaveW

Dave: sigh, I don't know about traceback... I've not had the easiest time with the server software (PyCS), so I'm not feeling enthusiastic about diving into it :( Maybe I'll switch to Syndicato sometime, it looks neat.

There hasn't been a lot of reaction on the mailing list. The Webware mailing list isn't that reactive. It's kind of sad -- there's just not a lot of enthusiasm or experimentation in the Webware community, for whatever reason. Maybe other frameworks have been able to steal away the developers who would be more interested in that sort of thing. Maybe all those developers just start their own projects (there's so damn many). I'll come back to it, but I already have a number of other projects which I'm involved in, so I'm spread a little thin, and I find it hard to keep my enthusiam going on Webware. I'm sure I'm not the only one. It's a problem. I've even had a hard time keeping my enthusiasm going with Web-SIG, which has the potential to deal with a lot of these same issues, and issues which I think are extremely important to Python.
# Ian Bicking

Ian,

No worries about trackback.

I tend to agree with your feelings about the status of a lot of python stuff. WebSig, Cheetah and Webware are good examples of projects that seem to be dying. SqlObject is looking much more positive. Unfortunately I can't find other active projects with architectures that I like as much as webware and cheetah.
# DaveW

To its credit Cheetah isn't really dying -- people seem to use it a fair amount. But no one who uses it is really modifying Cheetah itself. Webware has been similar in some ways. It's probably in large part on the basis of the original author's focus -- both Chuck and Tavis have not been focused on their software. I don't really know, but I imagine they are just focused on work, life, etc. So it goes. Making a piece of software that is developed in a lively fashion while the original developer is away can be difficult, and I don't know what the trick is. Open Source isn't nearly as easy as it's made out to be ;)
# Ian Bicking

Regarding Trackback and PyCS/PyDS: the problem here isn't the server software, it's more the site structure. PyDS generates no single posting pages like MoveableType or other systems do, it generates multiple postings pages (and archive pages) like Radio does. The problem here is that trackback autodiscovery based on HTML only works good if you have one posting per page, as the client software can't know which of those multiple postings to refer by what trackback URL. So PyDS doesn't generate autodicovery for trackback. But PyDS does add trackback:ping headers to the RSS feed, so aggregators should be able to get the trackback URL for individual postings from there. The trackback URL for entries is the same to ping as to see the trackbacks. So if you wan't to trackback one article, just use the "trackbacks" link and add that individually to your posting to ping. This might be possible with Typepad, but I don't know much about Typepad - but it is possible with moveable type, at least if you disable the autodiscovery stuff.
# Georg Bauer

Georg,

Yep I can enter the trackback url's into typepad manually, but they have not seemed to be picked up here. Anyway I can just use the comments instead. Thanks for the hints.
# DaveW

INRE WebWare enthusiasm....

For what it's worth, I picked up on WebWare for a project I'm working on, and I hope WebWare stays around for a while. I was impressed by installing it, getting it to work, and bootstrapping my own project with it in just a few days, having only a few minutes each day to tinker.

I plan on expanding the project to use Cheetah, once I get things sorted out.

I've continued to research other options since finding WebWare, but none of the other frameworks or application servers have made me reconsider.
# Mickey Hadicke