Re: Support for binary files, scalability andWindows port

Zbynek Winkler wrote:

> > I'll see if it's sufficiently easy to make the whole 
> > protocol and hashtree calculator depend on a couple template 
> > constants, and try to write it that way so we can wiggle it around and 
> > find a sweet spot.
> How is it going?
not bad. I would guess that I am about half way into the "non-prototype" 
version of this protocol. I would estimate another two weeks before it 
is really "operational" for VC use, after which I will experiment a bit 
and probably publish a version of monotone which has both it and the 
"conventional" queue/log/depot stuff, and we can try it out to see which 
is more pleasant to use.
so far I have found a nice networking library to use in place of 
boost::socket (netxx -- highly recommended), built and tested a basic 
binary packeting protocol, client and server i/o loops, and a dense, 
binary merkle-tree encoding which is templatized by bit fanout number, 
so uses much less space and network traffic. there's still work to do, 
but it seems to be proceeding well enough.
the "non-prototype" version of this protocol is a little more complex 
than the prototype, since it has to handle authentication and anonymous 
access, asymmetric sending / receiving roles (not just synchronizing), 
delta-encoded objects at the bottom of the tree, multiple types / 
collections of synchronized objects, and the possibility that its peer 
is trying to attack it. but it's not impossibly complex.
-graydon