ietf
[Top] [All Lists]

Re: A modest proposal

2013-01-22 04:00:47
----- Original Message -----
From: "William Jordan" <wjordan129(_at_)gmail(_dot_)com>
To: <ietf(_at_)ietf(_dot_)org>
Sent: Tuesday, January 22, 2013 4:57 AM
I've recent had to write a program to interface with a SIP lync server
and
in doing so have had to code to several rfcs.  After reading and
dealing
with implementation of the various rfcs I have read I have come up
with
what I consider "A modest proposal" to fix some of the problems I've
seen
with implementing a rfc.  I think anyone who writes a rfc should have
to
provide a working ANSI/C or GNU/C implementation of the rfc in
question.
 Specifically, I have worked with the SIP rfc (rfc 3261) and have come
to
the conclusion that whoever wrote the rfc has never coded a day in
their
life.  Whoever thought it was a good idea to allow multiple ways of
doing
the same exact thing would hopefully be deterred by actually writing
code
to do it.  I think a suitable punishment for those people would be to
write
each way of writing a from header on a blackboard 100 times... this
would
actually be less of the pain they've cause by making each writer of a
SIP
stack handle each possible way of doing things.

I find your proposal puzzling as many of those involved in the creation
of an RFC are experienced coders and do turn out implementations in
short order.  They know, they do not have to write the code, what is
easy to code and what is not.  If a choice is made, to have many
alternatives, which increases the complexity of implementation, then
that choice is made knowingly, in order to produce a protocol that will
better meet the requirements.

Code you write once, a successful implementation will execute a
countless number of times; what matters is that the protocol works every
time, not the length of the one off exercise to implement it beforehand.

If you had said that you had found flaws in the logic of the protocol
which could have been found by coding - and that does happen - then I
think that your argument would carry more weight.

Tom Petch

Anyways, that is my modest proposal, please respond or I will be
forced to
reply every day to this mailing list on each and every way the SIP
spec
sucks one email at a time.  FYI I'm not sure if GNU/C is the correct
acronym, maybe its POSIX/C.

Regards,
Bill



<Prev in Thread] Current Thread [Next in Thread>