[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Methods
- To: <nim@ops.ietf.org>
- Subject: Re: Methods
- From: "Joel M. Halpern" <joel@ivyplan.com>
- Date: Wed, 03 May 2000 11:15:06 -0400
- Delivery-date: Wed, 03 May 2000 08:15:59 -0700
- Envelope-to: nim-data@psg.com
Reading (and writing in) this discussion, I think we have agreement that we
need methods.
I have a guess as to what the question that is trying to be raised is.
Given that we are defining requirements for the modeling language,
what are the requirements on the language in terms of what it can say about
methods?
I will list some of the information we probably want:
Name
Signature (types of operands and results
description
Do we want formal or informal representation of error cases?
Do we want formal or informal descriptions of valid ranges on the
parameters and results?
Do we want formal or informal description of valid combinations of
parameters (if P1 is X, then P2 must be between Y and Z).
Clearly, we would like to be able to capture some of this
information. Equally clearly the more we require from the modeling
language the harder it gets to meet the requirements.
My personal take on this is to use whatever descriptive power I can get
from the modeling language we select, but beyond a reasonable method
signature I would not REQUIRE that the chosen modeling language have formal
capabilities for other things.
Once we have selected a modeling language, if it has additional
capabilities, we will have to decide collectively how to use them. And for
information we want to have that is not formally representable we will have
to decide where we want that to live.
Yours,
Joel M. Halpern