From: Mark Miller <>
Date: Mon, 25 Nov 2002 20:42:36 -0800
Subject: [e-lang] Serialization in Squeak-E (was: MAC calculation)

At 06:21 PM 11/26/2002 Tuesday, Rob Withers wrote:
>[...] Allowing active object graphs the facility to be boundless
>in a network, yet bounded by POLA capabilities, is very interesting to me.

That's very well put.  I think we can use that.  Thanks! rwithers12 

>It will be interesting to analyze, and a remote debugging capablity will be
>critical.  Whatever serialization we settle on :), we should look into
>whether we could exchange processes, lambdas, and contexts.

Funny you should mention that.  We just had some interesting victories for 
distributed debugging of E code.  I hope to explain more about this soon... markm rwithers12 

>> What's the current state of Squeak support for the Java serialization
>> format?
>heh, why would we want to do java when we can do squeak?   Everyone is a
>refugee here.  ;-)   sorry, there's no support for java serialization.

 From earlier correspondence I thought there was, but don't worry about it.  
I hope to drop Java serialization as well.   There's no way to get past the 
perception that it's Java-dependent, and no one is pushing it as a language 
neutral serialization standard. rwithers12

Have you looked at Tyler's Doc and how Tyler's WOMP uses it for 
serialization?   CapTP is not WOMP (they have different semantics), but I 
hope to switch CapTP's serialization to something at least close to WOMP's 
serialization.  Other serialization proposals are certainly welcome, so long 
as they have both an XML form and an efficient form.  (I know the OMG has 
something here I need to examine.) tyler rwithers12

>> Do you expect to use it for the initial CapTP experiments?
>I promise I will use all java serialization available in squeak: none.  (how
>does smart contracts deal with that assertion?)

It is compatible with all deployed smart contracting systems!  Ha, take that! rwithers12 

>  For the rest I will use
>Reference stream, which pickles a graph of objects, with a forward reference
> pass.    It wouldn't be that hard for E to support, would it?  :)

I have no idea.  If it's a serious suggestion, I'll try to take a look at 
it.   But my first question is, does it escape the 
Java-serialization-perception problem?  Will it be perceived to be Squeak or 
Smalltalk specific?  Is anyone pushing it as a language neutral standard? rwithers12

>It would be nice to have the serialization algorithm become part of the next
>rev of the startup negotition, wouldn't it?  XML, squeak reference, java
>serialization, term trees, ..

We definitely need to get the negotiation in there.  This negotiation should 
itself be textual, so binary-impaired participants can still play.   While 
there can be many surface syntaxes without much pain, it would be good to 
require no more than two: an XML-based one and an efficient one.  Above the 
surface syntax, I believe we will find we need to carefully define one 
abstract syntax (or, in x3c-speak, "infoset" or "document model") for the 
encoding and decoding of object graphs. rwithers12 and are my early drafts of  
a framework for organizing the examination of such matters, and of 
presenting term-trees partially within this framework. This page was written 
before I became aware of Tyler's Doc work, but I believe Doc should fit right 
in.  (I stupidly delayed announcing these pages until they were less rough, 
which never happened.) markm rwithers12

>Auditors is very interesting work.  I'd like to support those somehow.

Thanks!  Frankly, I don't know how to get to auditors from Squeak without 
some rather violent changes, but it's certainly worth a try! rwithers12 

Text by me above is hereby placed in the public domain


e-lang mailing list