[standards-jig] JNG Ramblings.

Mike Lin mikelin at MIT.EDU
Fri Aug 9 02:35:00 UTC 2002

> I second Matthias Wimmer's -1 to binary wire protocol. I know it would 
> make things increadibly easy for the server and programs, but difficult 
> for developers and admins.

I now have something that for the past year has been severely lacking in
all these discussions: code.

The Jabber XML Streams implementation I did in OCaml this summer is
about 600 lines of code in all, not counting the special XML parser I
had to write to make it work.

I wrote something to read my binary wire protocol and envelope format
yesterday afternoon. It's about 250 lines of code, and I'm a lot more
confident that it works right.

As I've been saying, I'm working hard to find the right blend so that we
get the benefits of the binary framing without sacrificing too much of
the protocol's elegance. The protocol even in raw form is still
perfectly readable in ethereal (there are just a few extra bytes spread
here and there), which I think matters much more than telnet-and-type.

Anyway, this is a question over which reasonable people can disagree. I
have thought for a long time, and I've become more convinced by
implementing it, that by using a binary protocol for framing we can get
enormous benefits for acceptable cost.


More information about the Standards mailing list