Re: CA protocol documentation.

[Brian Tieman <tieman@aps.anl.gov>]

Ralph,

You can apologize for the rough tone all you want, but I find this 
attitude towards documentation appaling.  That's not to say I think the 
documentation should be widely availiable--in fact, I think it should 
not be for the exact reasons you specify.

However, knowing that there is one and only one guru who knows what's 
going on scares the begeezes out of me.  As someone who implements Epics 
based systems--primarily with the PCAS, but with iocCore as well, I find 
it disconcerting every time a problem is posted to this list.  The first 
few responses--if any--are from well meaning, knowledgable, experienced 
users offering what amounts to an educated guess--it can't be much more 
than that because the internals of CA are unknown.  Then, perhaps a week 
later, the guru responds with the only answer I find I can trust.

I know Jeff is busy and I'm not saying he has to offer 24-7 tech 
support, but knowing he is the only guy who understands the CA 
protocol--the hing-pin of EPICS--makes me wonder what shape the 
community would be in were, God forbid, something to happen to him.  Is 
that the point where Epics ceases to exist?  Does someone step in to do 
their best only to have the "official" Epics CA libraries become buggy 
due to lack of knowledge/experience?  Do we then get an open protocol 
where people who need to can muck about?

Again, I'm not saying the protocol specification needs to be generally 
distributed, but I would sleep better at night if I knew it were written 
down somewhere and there was at least one other individual spending some 
amount of time developing with it.

Sorry for my tone, but I manage several projects on which I am the only 
person who knows how they work.  I've fought for years to try and get 
someone else involved to understand them in case I leave.  Having failed 
that, I have at least documented my efforts to the best of my ability 
such that someone else might have a fighting chance to take over after 
me--it's the only responsible thing to do.

Brian Tieman

Ralph Lange wrote:

>  >> I'm not sure if the information exists anywhere else 
>  >> but in the source code and with Jeff Hill. I don't recall
>
>  > That is what I thought at first place.
>
>  >> (over ten years) seeing a document that describes 
>  >> the internals of the protocol - unless Jeff alluded 
>  >> to it in some collaboration meeting.
>
>  > Is there any interest on such a document? Or a soruce code may be enough.
>
> There is some intention in this.
>
> Since the CA protocol is not well documented, everyone is forced to use
> the CA libraries that come with base. Thus we are able to develop and
> improve details of the network protocol without the overhead of an
> official protocol change.
>
> As soon as documentation exists, people will start writing their own
> libraries. In contrast to the original libraries these will be buggy,
> unstable and not compatible with older/newer versions of CA.
>
> Nevertheless, the authors will blame CA for this, flood tech talk with
> their questions and absorb valueable manpower to solve problems that
> wouldn't exist if they used the original libraries.
>
> Use the CA libraries from base. Don't start writing your own. Just don't.
>
> I think leaving the source code as the only documentation keeps the
> scare-off factor on an adequate level. We're open source, not open
> protocol.
>
> Cheers,
> Ralph
>
> ps. Sorry for the rough tone, but I think opening this would really be a
> serious danger. If CA gets suspected of being a flaky protocol, we're in
> trouble.
>