> firstly as most of you are no doubt aware 4 of the worlds 6 largest irc 
> networks are running p10 ircd's
> 
> would it not be easier to have a single codebase and have each network 
> #define out the features they don't require rather than everyone haveing to 
> make a patch work against a radically changed codebase

For many features, this is the mechanism used, though we prefer the
feature paradigm to #define's these days.  Only some of the more
radical features are left as separate patches--we are the _Undernet_
Coder Committee, after all, and experience has shown that admins
will activate features by accident or possibly design when we most
definitely do not want them to.

> secondly i have heard it mentioed that undernet are considering supporting 
> ipv6
> 
> the big danger with ipv6 is that with almost every end-site being allocated 
> a /48 regardless of size implementing a workable clone control policy will 
> be virtually impossible

Yep.  Good point, but IPv6 is coming, whether we want it to or not, and I
personally think it's a good thing.

> thirdly wy the hell does ircu 2.10.11 send a : before the last perameter of 
> a join or part message this is a waste of bandwidth as well as breaking 
> compatability with some clients

It only breaks compatibility with clients that fail to implement the RFC.
"Be conservative in what you send and liberal in what you receive."  Of
course, this principle works backwards, but the change was made originally
without realizing that older clients failed to implement the RFC correctly,
and we want to avoid making too many releases of the u2.10.11 code.

> and for that matter why does it send crlf instead of just lf on 
> server-server links this is also a waste of bandwidth and afaict no p10 
> implementation minds being sent just lf

RFC specifies "\r\n", the original code base does "\r\n", clients have a
right to expect "\r\n", and it simplifies buffer size calculations.
However, at some point I hope to change the server-server protocol to a
binary protocol.  This obviously would not extend to clients except
perhaps by a client-selectable extension in a far distant time and
probably another universe.
-- 
Kevin L. Mitchell <[EMAIL PROTECTED]>

Reply via email to