[rfc-i] diagram issues

Paul Kyzivat pkyzivat at alum.mit.edu
Thu Jun 26 11:29:48 PDT 2014

On 6/26/14 1:05 PM, Joe Touch wrote:
> On 6/26/2014 7:25 AM, Paul Kyzivat wrote:
>> Joe,
>> As an example of the point I am making, take a look at
>> draft-ietf-avtext-rtp-grouping-taxonomy-01. This started out with a UML
>> diagram - you can find it on the last page of:
>> http://www.ietf.org/proceedings/86/slides/slides-86-avtext-5.pdf
>> There is no way that this diagram could be reproduced in ascii art in a
>> draft. So the draft is mostly text. The diagram has been factored, and
>> bits and pieces of that done as ascii art. Those parts are helpful, but
>> the draft can't convey the big picture of how everything hangs together.
> I would say that the line drawing doesn't either, though.
> The diagram tries to convey several things at the same time
> (communication components, protocols, data types/structures), and it's
> not clear why it's useful to see all at the same time - except to
> underscore that the system architecture as a whole is in desperate need
> of clear interfaces and modularity.

Well, in this case the goal was to explain what exists. That is the 
first step towards improving it.

The fact that it conveys all those same things together is the point. It 
allows you to see how all the pieces relate to one another. Then you can 
focus on the points of interest to you.

I have found that such diagrams make it much more likely that you will 
discover previously unrecognized symmetries and silly redundancies. You 
don't easily get that by looking at the pieces individually - you then 
need to mentally build up your own conceptual model of how all the 
pieces fit together.

> Yes, there are corner cases like this, but do we really need to push the
> whole set of authors and documents to solving this case? This is a lot
> like the optimization of keeping enough source material to re-generate
> the doc *in case* there's a -bis update.

I don't think that cases like this are so unusual. It is just that you 
never discover them if you start with the assumption that you can't have 
a diagram more complex than you can draw with ascii art.


More information about the rfc-interest mailing list