[jdom-interest] Re: serialization Push toward a 1.0 API

[Joseph Bowbeer]

Ah, serialization....

----------
> Date: Tue, 09 Apr 2002 18:39:46 -0700
> From: Jason Hunter <jhunter at servlets.com>
> To: JDOM Interest <jdom-interest at jdom.org>
> Subject: [jdom-interest] Push toward a 1.0 API
>
> I think we're at a good point now where we can, with just a little push,
> define the JDOM 1.0 API.  There may be a little work after that point to
> have it fully implemented, and we're going to have to run through the
> JCP process, but I don't see any API issues outstanding that would keep
> us from defining the 1.0 API.
>
> I've setup in the TODO.txt file a new category at the top called, "API
> CHANGES FOR 1.0".  Below is the list as of today.  We start with the
> easier stuff.
>
> [...]
>
> * Make sure we have a story for serialization.  How much work is it
>   to support serialization across current and future JDOM versions?
>   See "serialVersionUID" thread especially Peter V. Gadjokov's
>   remarks at
>
>
http://lists.denveronline.net/lists/jdom-interest/2000-September/subject.htm
l
>
>   It may be OK to worry about fast short-term serialization only.
>
> Personally, I don't think we need to worry about long-term
> serialization.  That's what XML is for.  When you send XML over the wire
> and care about extensibility in the future, you should send it as XML
> and not as a Java-based serialized copy of the document.
>


----------
> Date: Tue, 9 Apr 2002 22:39:33 -0400
> To: JDOM Interest <jdom-interest at jdom.org>
> From: Elliotte Rusty Harold <elharo at metalab.unc.edu>
> Subject: [jdom-interest] serialization Push toward a 1.0 API
>
> I tend to agree in principal. However, I do think that if this is our
> philosophy we need to eat the whole hog; i.e. no JDOM classes
> implement Serializable, short term or long term. If the classes are
> going to implement Serializable, they should implement it correctly.
> --
>


Reminder: there was a lot more discussion about serialization in April,
2001.  (below)

http://www.servlets.com/archive/servlet/ReadMsg?msgId=7603

http://www.servlets.com/archive/servlet/ReadMsg?msgId=7646
http://www.servlets.com/archive/servlet/ReadMsg?msgId=7649
http://www.servlets.com/archive/servlet/ReadMsg?msgId=7655
http://www.servlets.com/archive/servlet/ReadMsg?msgId=7658
http://www.servlets.com/archive/servlet/ReadMsg?msgId=7670
http://www.servlets.com/archive/servlet/ReadMsg?msgId=7672
http://www.servlets.com/archive/servlet/ReadMsg?msgId=7673
http://www.servlets.com/archive/servlet/ReadMsg?msgId=7675
http://www.servlets.com/archive/servlet/ReadMsg?msgId=7684
http://www.servlets.com/archive/servlet/ReadMsg?msgId=7685
http://www.servlets.com/archive/servlet/ReadMsg?msgId=7687
http://www.servlets.com/archive/servlet/ReadMsg?msgId=7695

http://www.servlets.com/archive/servlet/ReadMsg?msgId=7697


Dennis Sosnoski pointed out that it's not just a choice between short-term
and long-term persistence, but also a focus on efficient transmission: "I'd
suggest instead that the focus should be on time and space efficient
serialization for transport. This is not one of the choices in your list,
but I believe it's the most important issue." (msgId=7685)

I raised issues related to subclassability (Serialiazable vs Externalizable)
and implementation-hiding (the need for a custom serialized format).
Implementation hiding is important even in short-term persistence if JDOM
documents are to be transmitted between different implementations of the
JDOM spec.  The default serialized format is very inefficient, though, so a
custom format would be desirable for reasons of efficiency alone.

Is XMLS a good starting point for a custom serialized form?

  http://www.sosnoski.com/opensrc/xmls/format.html

--
Joe Bowbeer