W3C home > Mailing lists > Public > www-http-ng-comments@w3.org > July to September 1998

Re: Thoughts on 19980710 drafts, do we need another object standard?

From: Ken MacLeod <ken@bitsko.slc.ut.us>
Date: 20 Aug 1998 16:46:03 -0500
To: www-http-ng-comments@w3.org
Message-ID: <m3lnojs4x0.fsf@biff.bitsko.slc.ut.us>
Mike_Spreitzer.PARC@xerox.com writes:

> The biggest sticking point is the divergence in type systems between
> CORBA and DCOM.  I can see two ways of coping with this: (1) create
> a biggish messyish type system that's as close to both CORBA's and
> DCOM's as possible, or (2) create one that is both simpler (at
> least, more orthogonal and uniform) and more powerful than CORBA's
> and DCOM's.  I can see pros and cons of both approaches.  I
> generally favor the latter.

We've been prototyping with BER compressed integers, mantissas, and
exponents for floats and the results look good so far.  On-the-wire,
we're using one type each of int, float, and opaque string as the
primitive types and then composites of dictionaries, lists, and the
primitives to build complex objects.  The format is self-describing,
and things like type declarations and object member names can be
reused by reference.

I'm not sure ``BER compressed'' is the correct term, that's the term
used in the Perl man page for the conversion, but I think it comes
from ASN.1 and ASN.1 doesn't seem to use ``BER'' in that exactly that
context.  In any case, what I mean is: integer values are encoded as
base-128 digits (7-bits/octet), high digit first, with the high-bit
set on all digits except the last.  This means the on-the-wire
encoding uses one type to serialize and implementations coerce and
range-check as necessary.

-- 
  Ken MacLeod
  ken@bitsko.slc.ut.us
Received on Thursday, 20 August 1998 17:52:24 EDT

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 13:07:28 EDT