[dcc2] Other comments

[myndzi]

> The question that comes into my mind is why you need to publish a port range ?
> Why should the remote party know the port ranges that you can listen on ? Or,
> in the other case, why should the remote party listen on a port range choosen
> by you ?

Perhaps you didn't read my other "big" post. The answer is, the remote
party needs to know what port ranges you are _willing_ to listen on.
This can be a subset of those you are actually _able_ to. The reason
for this is to resolve firewall issues better, so that both sides can
check their available outgoing ports versus a selection of incoming
ports and decide on one that works for both of them.
 
> But if one succeeds then the other could probably too :)
> Sending two files at once (with single-file transfer) is and will remain a
> common case.

Sure, but you may as well send them in serial. 

> Well.. I think that with the <name>[=<value>] pairs we can safely remove the
> restriction of the token order. The restriction would make the message
> parsing harder.

I didn't say everything should be ordered, but I just would like to
see the tokens split into the two disparate groups that they are. How
is it harder on parsing to assume that everything up to a certain
point are connection tokens, and the rest are application tokens (to
be passed on to the application using the connection)?

Have you actually tried to write something that complies to this crap yet? :P

-myndzi