James Tauber speculates on how Subversion (it's a version control system, used mainly by programmers) might be used to manage the code for his web publishing application, Leonardo, which powers this site. No updates, so maybe he has had no feedback.

I have nothing to offer regarding code in Subversion, but I have been using Subversion to manage the content, as opposed to the code, here. A few observations and speculations of my own follow.

will_be_able_to_do_all_the_things_i_want">Update: Leonardo will be able to do all the things I want


The Subversion book says to set up a project like this:

  • /branches
  • /tags
  • /trunk

So I did, when I put the content for this website under version control, but it makes no sense for my content. I think what we need for content is more like:

  • /drafts
  • /published/projects
  • /published/blog

At the moment I work on files in Leonardo's draft's directory and preview them through the web application. Leonardo works by mapping requests to text files in a wiki format, which it turns into HTML before serving them to you, so you can just give it text files in special places, or use a web interface to manage the files.

I edit the content from a Linux machine at home using XEmacs, and on the ISP's site via SSH using Emacs, and sometimes on a Windows machine at work using XEmacs. Subversion keeps all this in sync. If I make a change at home I can commit it, then update on the web server to see the change, but I could just as easily run a version of the website at home as well for content development.

When I'm ready to publish, I use a |svn move| to send a file to the blog or a permanent home in the site.


I'm happy enough with Leonardo + Subversion at the moment, but what else to I want?

I want to work on content in a variety of tools and formats for different jobs: plain text, 'wiki' text, html, and and/or Word for longer structured documents. I want to have a drafts area that is portable across different machines, (Subversion is perfect for this).I want to be able to browse my drafts as they will appear in HTML via a preview area on the website, as I can with Leonardo, but with more structure to the drafts area.

I also want to be able to have the same piece of content show up on more than one path in a site, for example a document like this could appear both in the blog 'stream' and as reference documentation in a projects area. The reference version of this content might then change, under version control of course, but I would likely leave the blogged version pretty much alone except for a pointer to the new version.

I want full-text search.

And a tagging system like, possibly with smart synchronization to my delic.ious.account, so that stuff I tag there propagates to my local system, and public resources tagged locally end up on automatically.

I want to be able to generate dynamic index pages automatically based on a query, say ask for a list of titles and summaries for all the blog entries tagged with

I want to be able to work locally on a draft, commit it and have the website automatically update its drafts area when I refresh the preview.

I want. I want. I want.

I'd better build it then, eh.


Managing revisions would probably be done by copying an existing published document back into drafts, revising it there, then re-publishing using a copy. In the case I mentioned above, where a piece of reference content might go in the weblog and under a timeless path, the new version would go to the reference path, and a the system should then be able to add a pointer to the old blog entry to show where there is a new version.

Bits of Subversion I think look promising for implementing the above include:

  • The ability to use properties to store stuff like renditions, or even the images that go with a page. So original content could be in wiki format, or a word processor, or HTML, or whatever, and a normalized rendition could be stored for each item. Although if I stick with Leonardo, it will just use directories to accomplish this.
  • Cheap copies of content, so publishing to multiple paths in a site becomes manageable.
  • The hooks that can fire-off when you commit in Subversion could be used to full-text index pages, and presumably this could be made smart enough that multiple copies of an item would not show up as multiple results. I would also like to be able to set up special indexes that would pull out content fraglets from documents and index them for retrieval on their own. These indexes could be specified using XPath or full-text (or XQuery I guess).
  • Properties could also be used for content tagging. I don't think you can query on properties using Subversion (eg ask 'list all the docs with the XHTML tag'), but the full-text indexer could take care of that kind of index as well. Outside resources such as web pages would need a stub-record so they could have properties attached.

$LastChangedDate: 2004-12-22 07:23:05 -0600 (Wed, 22 Dec 2004) $

Site version: $Rev: 144 $


comments powered by Disqus