Re: [Boost-users] [serialisation]Portable Binary Archives

From: Cliff Green (cliffg_at_[hidden])
Date: 2007-06-27 12:21:39

>> ... endian library 

> It looks as though it hasn't been proposed for review:
> http://www.boost.org/more/formal_review_schedule.html
> 
> ... quite a bit of discussion about it, with many 
>interesting
> ideas/additions/alternatives raised, so a mini-review 
>(or something)
> would probably happen.

It's been a while (almost a year?) since Beman uploaded 
his endian utility. Beman, is endian-06.zip the latest, in 
the vault?

Is the first step in creating a portable binary archive 
collecting and summarizing the discussions from a year ago 
and performing a mini-review to make the endian utility 
part of Boost? This would provide endian utility for 
integral types. The floating point utilities from Johan 
(which I haven't looked at yet) would need a place to 
reside (they could reside in the serialization archive, 
but it seems natural to have them as a separate utility).

I'm going to look at the example serialization code - one 
concern I have (at this point an uneducated concern, since 
so far I've only casually read the serialization docs) is 
that I'd like a way to use a binary archive that doesn't 
have the "metadata" that is normally provided in 
serialization archives - I want to be able to read / write 
or send / receive buffers of packed binary data where I 
have complete control over every byte - I don't want type 
ids, version numbers, pointer sharing semantics, etc 
(unless I explicitly put them in my code). I realize this 
is an orthogonal concern to the portable binary mechanics, 
but I wouldn't be surprised if the users and applications 
wanting compact and efficient binary archives overlap 
significantly with the users and apps wanting control of 
every byte in the stream.

Cliff