WebLogs and RSS, James Robertson, Cincom
A blog (web-log) is an on-line journal. To James, the point of a blog is that it does not have to go through the marketing department. Other companies (Oracle, even Sun but not IBM yet) use blogs for this purpose. James' and Dave's code is in public repository. Blogs are becoming common on technology, politics, hobbies. The Blogosphere is also where everything discovered 20 years ago is being reinvented, sometimes less optimally.
James' wrote a blog program as a demo of the VWwebtoolkit. His first thought was 'How hard can this be?' Gradually he discovered more and more things that a blog can do (or not, as it chooses). He soon discovered that he needed to edit misspelled posts, that it takes a long time to get readers, that he was the only one posting. Blogs are unlike Wikis in that they tend to have a single voice and to be massively cross-linked.
Most blogging systems store stuff as HTML and archive them in separate directories, fast to load but slow to edit. James stores stuff as BOSS. BOSS schema migration takes care of shape-changing.
(I remarked that, given the way the code works, having a daily BOSS file influences how much you see when you go to an archive post; some blogs move the reader to a old post within that week or fortnight of posts, which can be helpful for seeing its context, or just for tempting them to read what you posted the day before or the day after. When I looked at this, I altered the code to make the BOSS increment configurable.)
CSS was a mystery to him at first so he laid out pages in tables. This was wrong. He now uses a template.
He has a basic BlogEntry object that has evolved to handle comments, etc. When he started, he did not know VWwebtoolkit servlets too well. He should have used a get servlet but started with page Smalltalk code which grew and grew. His strongly advises to look at servlets first and avoid ssp pages until you find you need them.
At first he had ~5 readers per day. Then he got email asking him to add an RSS feed, something he'd never heard of. He learned (that RSS has its own politics so RSS formats have some surprisingly optional tags and there are modules that duplicate these optional tags, etc., etc., all making life harder for BottomFeeder and similar newsreaders :-)). RSS is a syndication format, needed because while some people go to a blog's web page and see if it has new stuff, others have a news aggregator that gives them a summary of which sites have changed and how much. At first, he added an RSS feed by building an XML document by adding tags, which is the stupid way to do it. Later, someone told him about SAX drivers and he used one. He wrote a small bit of code to parse RSS off the web and put it in the repository. Then Dave wrote a UI for it and so BottomFeeder was born.
(www.hebig.org/blogs/archives/main/000877.php has some aggregators.)
BottomFeeder started with an RSSFeed and RSSItem, shown in a standard 3-pane screen. Handling all RSS formats from 0.9 to 2.0 is hard; they're not as alike as the version numbers suggest. You have to handle lots of internet errors, lost sites, garbage XML, network errors. Storing the XML as a file took 60 secs to load when feeds rose to hundreds so he moved to BOSS. They use TwoFlower as their HTML browser (may soon move to Michael Lucas-Smith's browser). BottomFeeder started as one package of 4 classes and is now 21 packages and 120 classes. Jim (Columbia DC) and Dave (Vancouver) coded, Rich Diemers (Minnesota) created the documentation, Holger Klien (in Germany) helped with TwoFlower. They used IRC a lot.
A blog increases its exposure by registering with various sites. You can also support trackbacks and pingbacks. In the blog world, XML-RPC is what people use, not SOAP. Roger Whitney (San Diego) did an XML-RPC VW app which Jim ported from 3.0 to VW7 and made work with servlets.
The other pattern people use is REST. It is simple, mostly just a URL, but sadly, few people use it. Pingback and Trackback were different (former was autodiscovery) but now they are both autodiscovery and there are four distinct but same-name-using forms of trackback (HTTP Get, HTTP post which are complex and bad, CommentAPI which is better, and a fourth).
The web toolkit had no way of doing REST as it did not expect anyone to reply with raw content added to the middle of the post, so he had to write a method #rawContent to do it. It is a multi-path method only because he wanted it to work with several versions of the toolkit. On any one version, it will only go down one route.
Cincom use RSS feeds for bug tracking, the internal and public Store DBs, the Wiki,... Some Microsoft projects use blogs for internal communication.
Previous Topic
