Re: Project for SEUL?

[Ben_Tilly@trepp.com]

> Hi Ben,
>
> Well I can't say that the LinuxKB is doing that, since we're a web based
> documentation engine rather than a prog on the users system.  I do think
> that there is definately a place for something like this, though I think
> it should not only be able to return results for the included
> documentation on your harddisk, but also query an online DB like the
> LinuxKB.  I know Cheek Consulting's KB has an app that you can install
> which will query their database for answers, but I don't know how much
> it's being utilized by people or what it can do feature-wise.  (Jospeh
> Cheek is on the linuxkb-discuss list so maybe he can jump in here and
> provide some stats & info.)

I wish that Doug left my original description in.  Everyone's immediate
impression is something very different (and more complex) than what I was
thinking. :-(  I am talking just a local, well-trained, infobot to be
friends with you and help you through some of those, "I know there is an
easy answer somewhere but I don't what to begin to ask" situations.

> Anyways, IMHO for something like this to really work you need:
>
> 1) The distributions to include this on their CD's

Agreed, important.

> 2) The ability to deal with man, texinfo, and /usr/doc files

The approach I was thinking would include references to web pages, perldoc,
newsgroups, IRC channels, system configuration files, paper books and
anything else that people asking questions would get directed towards.

> 3) Ability to query an online DB like the LinuxKB if internet access is
> available for additional content.

Wouldn't need that, although updating of the knowledge should be easy.

> 4) Get people (newbies) aware of the app

Agreed, important.

> 5) Bonus points for phrase searching, rating "hits", and natural language
> interface.

Infobots on IRC channels already do a good job with this.

> 6) A nice interface (use Lynx or something)
>
Interface is that you ask a stupid question.  If you ask anything
reasonably simple and relatively common in the  way that confused people
tend to ask the question, you will probably get a helpful answer.  It is
not the, "Here is your definitive reference" type of thing, instead it is
more like, "Before you do any work, ask this, it is worth a shot."

> #1 is easy.  I'm sure Roger and I have enough contacts for Red Hat, SuSE,
> and Debian to get this to happen.
>
:-)

> #3 is easy, because I can make it happen, and having an OpenAPI for the KB
> is one of our future goals for our site.
>
Irrelevant for what I was thinking about.

> #2 is probably very doable as well assuming your fluent in Perl.
>
Right now "awake and coherent" are bigger barriers than Perl...I was
mentioning this to Doug as a lower priority item for me that I thought
someone else might like to do before I get around to it (if ever).

> #4 is going to be a little difficult.  If we can convince the distros to
> add a message to their install sequence which says "run command x" for
> help (can't understand why they don't do this already) your golden.
>
If it is really helpful I suspect that this will take care of itself.  The
critical point would be if commercial programs began to submit databases
based on level 1 questions that show up in their help lines.

> #5 is going to be very difficult without using an existing search engine
> like ht://dig and then extending it.  Problem with this is disk space
> requirements go way up, and people aren't likely to install a 100MB help
> tool on their system.  This is why interfacing to a site on the web is so
> useful- we've got all the resources to deal with this sort of thing
> already.  Of course they've already have to have internet access for this
> to work. :(
>
My concern is startup time.  Size should not be an issue.  (Partly because
the advice consists of short tips and the human request interface is very
stylized.)  Luckily the development of this interface is already done.
Infobots have been around for a few years and have proven quite successful.

> Anyways, if you or anyone else is interested in working with us to develop
> such a tool to interface with our site, I'd love to hear about it.  We
> don't have any people resources to help in the development of this right
> now, but I'd be willing to talk to you about our OpenAPI which hasn't been
> spec'd out yet.  Hell, we don't even have a cool name for it yet :)

Hmmm...

KNowledge Online via the WWW.  KNOW for short.

Nah, too silly.

> If you're interested, I suggest we take this to the linuxkb-discuss list.
> Subscription info can be found on www.linuxkb.org.

I don't think that I will do this for 2 reasons:

1)  I am not tackling this project in the near future.

2)  I had a lazier approach in mind than trying to generate the knowledge
base myself.

Cheers,
Ben