Some questions on the draft HTTP document

A few of us have been looking at building an HTTP protocol stack for
ILU; that is, allow ILU methods to be called on ILU objects via HTTP,
and allow HTTP servers to be perceived as ILU objects.  ILU uses an
idealized RPC protocol for this purpose, and at first glance, it seems
that HTTP could support this RPC protocol quite nicely (if somewhat
expensively).

However, after looking at
http://www.ics.uci.edu/pub/ietf/http/draft-ietf-http-v10-spec-00.txt,
it's not clear what the right way to pass parameters is.  For example,
suppose I had an object which supported the method
"BoundedCalculate(Calculation,TimeLimit)", where Calculation is a
string, and TimeLimit is an integer.  The meaning of BoundedCalculate is
``run the BASIC computation described in Calculation, using a CPU time
limit of TimeLimit units, and return the result (a real number),
signalling the exception InsufficientTime if the computation cannot be
finished in TimeLimit units of CPU time''.  What would be the correct
mapping of an invocation of this method to HTTP?

First guess (assuming that `http://foo.bar.com/calcservice' is the URL
of the object):

	BoundedCalculate calcservice HTTP/1.0
	Content-Type:  message/multipart, boundary="-------"

	-------
	Content-Type:  application/x-ilu-string

	(Some string here, presumably valid BASIC)
	-------
	Content-Type:  application/x-ilu-real

	4.5
	-------
	
for the request, and what for the response?

Perhaps more significantly, what is the future purpose of HTTP?  If it
is to mainly become a transport for supporting the 8 `special' methods,
perhaps we should avoid trying to use it.

Bill

Received on Tuesday, 6 June 1995 18:21:57 UTC