Mailing-List: contact poi-dev-help@jakarta.apache.org; run by ezmlm
Date: Tue, 22 Jul 2003 12:20:02 -0700
From: Chris Nokleberg <chris@sixlegs.com>
To: POI Developers List <poi-dev@jakarta.apache.org>
Subject: Re: Improving POIFS performance
Message-ID: <20030722192002.GC11770@fez.sixlegs.com>
References: <20030722180851.GA11696@fez.sixlegs.com>
 <00a001c35081$0b234550$0d8fcda3@tdryan>
 <20030722185027.GA11770@fez.sixlegs.com>
 <00b601c35084$91221160$0d8fcda3@tdryan>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <00b601c35084$91221160$0d8fcda3@tdryan>
User-Agent: Mutt/1.4i
Status: O

On Tue, Jul 22, 2003 at 03:07:58PM -0400, Ryan Ackley wrote:
> > Sorry, it's not likely to be open-sourced in the near future, but I'll
> > let you know if the situation changes.
> >
> > Chris
> 
> Do you mind if I ask how you guys did it? Did you simply reverse engineer
> it? Or did you have a spec?

Unfortunately there's very limited public information available for
PowerPoint, much less than Excel. Maybe the situation is the same for
Word? There were some specs on an old MSDN CD, but they're only useful
up to a point. Basically they just describe each record, but not any of
the interactions between records.

So in the end it came down to a lot of reverse engineering--dump out the
record tree, make a change, dump again and diff. The hardest part is
that when writing out a file, if you get anything wrong PowerPoint just
refuses to open it. I'm sure this all sounds familiar...

Chris