Abstract Away the Fetching Process

The screenshot in Torsten's blog shows him using RSS Bandit to monitor his Windows Event Log. Now I thought this was particularly interesting because I wrote most of the code that does the processing of RSS feeds and beside a tweak or two by Torsten to handle local file:// URLs there were no changes that would support such functionality. I thought about it a little bit and decided that Torsten must have written an app that monitors his event log then writes the current state at periodic intervals to a local RSS file which Torsten is subscribed to. Of course, I might be wrong but since Torsten is on vacation he isn't here to correct me so I just hope my assumption is right.

My mind started to race because I thought a number of things simultaneously on digesting the screenshot

1. As long as information can be identified (preferrably as a URI) and represented as an XML infoset there is no reason why an RSS information aggregator shouldn't be able to subscribe to it.
   
   
2. The user should be shielded from the differences between "actual" XML files fetched from the web via HTTP and "virtual" XML retrieved by non-traditional means.
   
   
3. The same way some Windows apps today can consume data from any ODBC data source (e.g. Crystal Reports) it would be interesting if the same could be done with RSS data sources or even more generally XML data sources.

I'm not sure how far I'll take this idea or even if I'll start working on it but I will at least make sure that the the RSS processing code abstracts away everything that has to do with how the information is fetched from the GUI. This should already be the case modulo some try...catch blocks for WebException but I need to make sure.

Besides the engineering problems (hah, coders are so pretentious) there are also some usability issues that need to be solved. Currently when wants to subscribe to a new feed there is a dialog box where the user enters the name of the feed, its HTTP or file URL, and what category it should be placed in. Now how do I make this model work in a world where the user can subscribe to their machine's event log or the contents of a network share via some "RSS data provider"? The only thing that comes to mind is using custom URI formats but I somewhat suspect that the web architecture weenies like Tim Bray would shit a brick. Interesting stuff.

Abstract Away the Caching Process

I've mentioned in the past that I'd like to add an abstract caching layer to RSS Bandit. Currently my caching code is all hard wired to assume the cache is on the file system but the closer we get to shipping Yukon the more I want that at least on my machine I use it as my cache. Abstracting away the cache seems fairly straightforward for my current implementation, all I do is load and save files (somewhat analogous to HTTP GET & PUT). Creating a CacheManager that supports both these operations then creating specific implementations like FileCacheManager & YukonCacheManager seems like the thing to do.

The complexity comes in when I want to do more with the cache though. I guess I'll cross that bridge when I get to it.