Even though I was talking about using a"bot", maybe we don't need that much versatility.You could do it with AJAX or just server side (boring). ;P
I'm think about a client-side program to reduce bandwidth. The program (for instance, for the purposes of discussion, let's say a java applet) would just gather information, search several databases, and return an organized response. This would have at least two advantages:
1.) Gives relevant information to the potential poster(related, solved, posts; howtos/guides; etc...) These would be returned to the screen in categories. One box containing links to the Howtos, another with links to related, solved posts, and so on.
2.) Gets relevant information from the user. Depending on the nature of the problem, certain standard information would be requested. For instance, if it were a hardware problem, the user would be asked to add the output of lspci and dmesg and lsmod and so on, to their post.
Both of these don't really require a "bot".
Info from user
Getting the information from the user is trivial; Basic flowchart logic gives a series of finite problem types. Each problem type has a corresponding set of information to be requested from the user.
Info to user
Returning the information is also not that hard. After being asked a series of questions (another flowchart) our program has a bunch of keywords (a bit more than that since our program gets information in response to specific questions, our program has a correlation between the two. This correlation allows the program to search more intelligently--such as what sites to search at, for instance).
One of the key open questions is: What are the different technologies currently used by the different Ubuntu websites/tools? Database, Search feature, server side programs, client-side programs, are the different sites connected in some ways? How hard would it be to have a unified search feature for the different websites?
On this last question, if there were a unified search feature, and there were plenty of meta-data on the different sites, then this would improve things dramatically. For instance, instead of using seperate search engines, our program would just use one. When the data is returned, our program analyses the meta-data to extract basic "type" information such as "is this a howto" "is this a post" "is this a recommendation" etc..etc.. This will allow our program to return an organized reponse to the user.
Any ideas on how we can find out what website/webserver technologies are being used on the Ubuntu sites?