From: Robert Ramey (ramey_at_[hidden])
Date: 2003-09-14 14:24:15
> BTW, I look forward to a description of the requirements for the creation of
> new archives. I believe the new design very nicely supports user written
> archives and in my experience this becomes a very important capability (your
> requirement #8). As an example, on a recent project we had to write an
> 'archive' (we split archives and call them 'Reader' and 'Writer' classes)
> because we needed to support certain binary byte orderings for sending /
> receiving messages. We use this type of archive on the boundary of the system
> and then used o/istream-based archives for communications within the system.
> All the while, the message objects are oblivious to these different formats.
We're on the same page here. I didn't think it was appreciated
how important this requirement is. From my point of view its not
so much for practical reasons such as those you've mentioned,
though they are very important,
but rather the impossibilty for me to keep in my head the
requirements of all possible combinations for serializations
and archives. Its really necessary to minimize the the
"conceptual surface area" required to understand something
like this by keeping it in orthogonal composable pieces. Its
already way past the point where I can understand more than
one aspect at at a time.
draft #12 contains this new section of the manual - its already
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk