Re: [boost] Re:New mailing list: boost-serialize

From: David Abrahams (dave_at_[hidden])
Date: 2002-12-10 11:22:37

Robert Ramey <ramey_at_[hidden]> writes:

> What is the motivation for having discussion regarding serialization
> and/or persistence take place on a separate list?

Mostly that it is a very specific design discussion that might easily
overwhelm the already heavy traffic on the main boost list.  But I am
not rigid about this idea.  If more people prefer to have it here,
that's fine.

> Is it inherently different in some way than other boost projects?

Different from /some/ other boost projects, yes. Several other boost
projects have their own lists: Boost.Python, Spirit, and Boost.Build
come to mind.

> Has it consumed more "space" than say shared_ptr ?  Is is so
> disjoint from other boost topics? Is the discussion some how
> different than other boost discussions?.  Are other topics
> candidates for a separate list?  Would this "other list" be
> searchable with the WWW interface?

I guess not:
http://sourceforge.net/mailarchive/forum.php?forum_id=17460

> I concede that I can't follow every detail of the mailing lists.  I
> really only place close attention to those topics that I already
> have a personal interest in and/or "catch my eye" for some reason.
> But I can't see that the serialization discussion was so much
> different than others that take place on the list.

That's fine, your vote definitely has weight in this matter.  If you
have a strong preference for doing it here, I have no objection.
Maybe that would be best.

> Rather than commenting on the details of your review - which would
> only repeat points previously made by myself and others - I would
> note that it re-enforces my skeptism that a consensus can be
> achieved on an implementable set of features.

I wonder why? Most points were of the form "please give careful
consideration to ..." There were only a few points on which I set out a very
strong requirement:

    * Registration of participating classes must not be required to be
      monolithic.  More generally, the library must support users who
      use polymorphism to insulate themselves from compilation
      dependencies.
 
    * ...Avoiding virtual function calls for serializing large arrays
      of small objects (e.g. complex or rational numbers) must be
      possible

-- 
                       David Abrahams
   dave_at_[hidden] * http://www.boost-consulting.com
Boost support, enhancements, training, and commercial distribution