Re: [Issue, March 20th 2002] Serializer vs. De-serializer

Forwarded copy for the comments archive.  Asir:  if you have a problem with
this resolution, please let us know.  Thank you.

------------------------------------------------------------------
Noah Mendelsohn                              Voice: 1-617-693-4036
IBM Corporation                                Fax: 1-617-693-8676
One Rogers Street
Cambridge, MA 02142
------------------------------------------------------------------



----- Forwarded by Noah Mendelsohn/Cambridge/IBM on 04/24/02 04:01 PM -----

                      Noah Mendelsohn

                                               To:
<asirv@webmethods.com>

                      04/19/02 08:44 PM        cc:      xml-dist-app@w3.org

                                               Subject: Re: [Issue, March
20th 2002] Serializer vs. De-serializer(Document link: Noah
                                               Mendelsohn)







Asir:  on last Wednesday's call, I was asked by the workgroup to prepare a
response to your comments regarding the perspective from which the
encoding is presented.  One of the comments you make is:

   "I prefer rules for both serializing and
    de-serializing. If that is not a possibility,
    then I prefer rules for serializing as
    opposed to de-serializing."

As you know, the text in question traces to a rework that Gudge and I
undertook following the meetings in France.  The pertinent parts of the
resulting specification say:

   "The encodings are described below from the
    perspective of a de-serializer. In each case,
    the presence of an XML serialization is presumed,
    and the mapping to a corresponding graph is
    described.

    When serializing, more than one encoding is
    typically possible for a given graph.
    For example, the order in which multi-reference
    graph nodes are serialized is generally
    insignificant. When serializing a graph for
    transmission inside a SOAP message any
    representation that deserializes to the
    identical graph MAY be used; when receiving
    an encoded SOAP message, all representations
    MUST be accepted. "

There are two aspects to your question, so let me deal with them
separately.

(1) I hope the above makes clear that we have in fact formally specified
the rules for both encoding and decoding.  The second paragraph quoted
gives the rules for encoding, albeit by making reference to the rules for
decoding.

(2) There is a stylistic question as to whether we might have approached
the presentation in the opposite manner.   I had the same intuition as you
(lead with the encodings), but when I sat down to do the writing I found
it was extremely difficult.  Gudge read a draft that I wrote, also felt it
would be better the "encodings first" way, set out to change it, and
reported the same experience as I had had: it was much more difficult.  He
changed his mind.   In short, I
believe this traces to the one-to-many correspondence between a particular
graph and its many equivalent encodings.

So, I don't know whether you'll find that answer appropriate or
reassuring, but it does reflect our experience, and the workgroup asked me
to communicate that to you.

You also lists a number of section headings with titles that refer to
"encoding...".  As noted in (1) above, I believe we are in fact describing
encodings as well as decodings. The two are duals of each other, and are
presented as such.   For these reasons, the workgroup is proposing to
leave the text as it is.  Although no formal issue has been opened, I'm
sure the workgroup would appreciate your letting us know whether this is
an acceptable resolution.

------------------------------------------------------------------
Noah Mendelsohn                              Voice: 1-617-693-4036
IBM Corporation                                Fax: 1-617-693-8676
One Rogers Street
Cambridge, MA 02142
------------------------------------------------------------------







"Asir S Vedamuthu" <asirv@webmethods.com>
Sent by: xml-dist-app-request@w3.org
04/16/2002 07:55 AM
Please respond to asirv


        To:     <xml-dist-app@w3.org>
        cc:     (bcc: Noah Mendelsohn/Cambridge/IBM)
        Subject:        [Issue, March 20th 2002] Serializer vs.
De-serializer


I raised this issue on March 20th, 2002 [1]. Original text is,

"[Comment B] Section 3.1 says that rules are described "from the
perspective
of a de-serializer".  But,

- title for section 3.1 is 'Rules for Encoding'
- sub section titles are 'Encoding ...
- section 3.1.1 is a rule for serializing an edge
- section 3.1.3 has rules for serializing generic, struct and array
- section 3.1.6 has rules to serialize the value of arraySize
- Appendix A has rules for serializing app defined names

BTW, in an ideal situation, I prefer rules for both serializing and
de-serializing. If that is not a possibility, then I prefer rules for
serializing as opposed to de-serializing."

[1] http://lists.w3.org/Archives/Public/xml-dist-app/2002Mar/0193.html

Regards,

Asir S Vedamuthu

webMethods, Inc.
703-460-2513 or asirv@webmethods.com
http://www.webmethods.com/

Received on Thursday, 25 April 2002 09:40:07 UTC