[rfc-i] RFC Heresies

Tim Bray tbray at textuality.com
Tue Mar 27 22:23:56 PDT 2012


That sounds like a really good BOF, wish I could have come.  There
were two items stated as conventional wisdom that I think are really
wrong, so let me convince everyone that I’m a crazed, drooling madman.

1. No, we don’t need graphics

I’ve thought for years that one of the great strengths of the RFC
series was the absence of structure tables and flowcharts and various
other visual expressions of architecture astronautics; forcing the
authors to explain Internet protocols in simple clear linear English.
At the end of the day the audience is coders who have to express
behavior of network elements in a linear fashion, and I don’t know any
way of programming computers that involves showing pictures to them.
I think the potential harm from letting graphics in the door greatly
exceeds the potential benefits. It’s a slippery slope that leads
inexorably to UML. The one exception that seems plainly obvious is
math, and I guess you do need some of that to design the Internet. But
that’s the only exception I’d make.  Plus ASCII art for ladder
diagrams.

2. No, we don’t need metadata

YAGNI.  Really.  Requiring metadata significantly increases the
cognitive burden on authors without conferring any particular benefit
on readers.  This argument has been going on for years and the
proponents of metadata always talk about the magical wonderful things
that will become possible once everything’s tagged and structured, and
we never get to eat that cake. RDF died for a reason. The Semantic Web
oops I mean “linked data” is languishing for a reason.  I find the
existing corpus of RFCs to be nicely searchable as it stands.


More information about the rfc-interest mailing list