In message
<C1256A00(_dot_)004562BD(_dot_)00(_at_)notes06(_dot_)telefonica-data(_dot_)com>,
joaquin.riveraro
driguez(_at_)telefonica-data(_dot_)com typed:
< Perhaps we ( the IETF ) should have a library of standard,
<downloadable translation / formatting tools that would help people to write
<in whatever format they choose, then convert it to the required ASCII.
<However, this would still not solve the problem os ASCII's poor diagram
<capability.
I am sure that will help, while the discussion on the standard format goes
on,
the tools will be helpfull to everyone whatever the final decision should be.
there is no substitute for good graphics design skills/ability -
havign said that, some tools WOULD be nice - i think its irrelevant
whether the tools render the output as GIFs or PDF or ascii - the
problem some people appear to have is focusing.
in practice ,there's 3 or 4 diagram types:
1/ packet headers- here the conventions used in rfc791 onwards are
EXCELLENT since they are cpu agnostic- since they are also labelled
they are no more national language specific than a program is:-)
(e.g. C structure or Java )
2/ state machines - these are not too bad - yo ucan use the same
approach as is used in old 60s/70s flow charting/call graphing in
general, quite clearly....the most complex state machine (e.g. new PIM
SM spec, or TCP) are not too hard
3/ packet exchange examples (e.g. time sequence diagrams) - i think
these are trivial (except occasionally in multicast:-)
a tool for these would be pretty simple to build...
(something could back end off of emacs, powerpoint animations, ns
animations and magicpoint etc)
4/ topology based expostion (i.e. routing protocols) - these are
generally very hard - ascii makes you think a LOT, as i said before
about keeping the examples simple
any other?
so how about a project to develope some tools for the last trickier case
above? (btw, i dont see how XML helps one bit - PDF or PS are the only
options for platform independnt rendering, and even then there are
problems with portability and fidelity) - and specifying the actual
editing/wordprocessing toolset is not on!
cheers
jon
p.s. how mayny people really read a protocol spec on a PDA? i mean the
time i do it is when coding, and when coding i want the spec in a
window, the code in a window, gdb in another window, tcpdump in 2 more
- seriously.