Hi,
The design of the IAC is still in the works - you can catch up some
information regarding the design options on the dev group (tptp platform-dev
mailing list).
I will follow up regarding the status of the prelim. design document.
Thanks
Guru
"Musachy Barroso" <musachy@gmail.com> wrote in message
news:dg50e1$e5j$1@news.eclipse.org...
> Hi
>
> Is there any documentation explaining how IAC is going to work? if there
> isn't where can I find the new implementation in CVS?
>
> thanks
>
> musachy
Hi, everybody. I'm doing a project that requires me to take code from an
eclipse environment, put it on a remote server, compile and run it
there, and gather output/logfiles and performance data. From what I've
read, tptp sounds like the perfect tool for me to use and even
contribute to but I'm having a terrible time finding documentation on
how to write agents and how to communicate with them.
Can somebody point me in the right direction? Many of you seem to know
how to use tptp, how did you learn it?
At the moment we're in a transitional phase with regard to how to implement
agents. We have two Agent Controllers with different agent development
models. Which you will want to work with depends on your schedule
constraints and how much of the workbench TPTP code you need to interoperate
with.
The older Agent Controller (which is simply listed as Agent Controller on
the downloads page) is fully integrated with all the workbench features and
has been tested with them. The newer Agent Controller (which is listed as
"New Technology Agent Controller" on the latest 4.1 downloads page or
"Technology Preview" on older download pages) is fully functional and
supported by a communications API on the workbench, but isn't tested with
the other workbench features yet.
The older Agent Controller will be phased out over the next year or two, but
if you want something today, you might want to consider it. You can find a
document on how to write agents for that Agent Controller at this link:
If you are interested in working with the newer Agent Controller, there's an
overview in the October issue of Dr. Dobb's Journal which you can find at
this link:
Andy Kaylor wrote:
> This should give you enough of a start to ask some follow-up questions.
>
> -Andy
Hello Andy, thank you for the response. That information about the
transition stage cleared up a lot of my disorientation but not all of
it; as you predicted, I have more questions. :)
I had a book of magic tricks when I was a child. It had instructions for
lots of really cool tricks but most of the instructions were on the
order of "Make the coin disappear in the usual way, flourish your hands
and wait for applause."
Believe it or not, I had already found and read all of the documents you
just posted but I found them to be much like that book of magic tricks;
lots about what tptp could do without much in the way of how to get the
platform to do it. Is there anything in the way of a walk through to
create a hello-world example? High-level documents like that may be
useful for someone with lots of tptp familiarity but they're quite
frustrating to those of us trying to *build* that familiarity.
I'm afraid we don't have any documentation of that sort. As you probably
know, we do have some samples, and you could probably find someone here to
explain how the magic works at any given point.
The "samples" for the old Agent Controller are actually better described as
"exemplary tools" in that they do a lot of tool-specific things that make
them more complicated than you would want if you're just trying to figure
out how this stuff works. If you want to go that way, I have a very basic
sample that I wrote when I was trying to figure out how that stuff works.
It's more on the order of "Hello World" and I'd be happy to send it to you.
The samples for the new Agent Controller are much simpler. Our sample
collector, for instance, only measures how long you let it run. If you
download the SDK package for the new technology Agent Controller, you will
get the samples.