Using Finite state machine for Protocol design

[Shashank Khanvilkar]

> Hi,

> Will really appreciate for any comments on the below:
>
> Are there any Open Source tools that can convert a protocol design, as a
> Finite State Machine,
> directly to C code. (I can, of course, google for it, but I would like to
> get comments from people who
> have used such tools and how effective they are).

To ADD to my previous mail:
I have also found out in many codes, that they use function pointers for
different types of messages. Thus many people define a structure for
expected messages, which has a pointer to a function which deals with the
particular message, which , in my opinion is pretty neat.
Comments?

>
> --
> Regards
> Shashank
> [url]http://mia.ece.uic.edu/~papers[/url]
>
>

[Michael B Allen]

On Thu, 31 Jul 2003 12:34:37 -0400, Shashank Khanvilkar wrote:

>> Hi,
>> Will really appreciate for any comments on the below:
>>
>> Are there any Open Source tools that can convert a protocol design, as
>> a Finite State Machine,
>> directly to C code. (I can, of course, google for it, but I would like
>> to get comments from people who
>> have used such tools and how effective they are).

> To ADD to my previous mail:
> I have also found out in many codes, that they use function pointers for
> different types of messages. Thus many people define a structure for
> expected messages, which has a pointer to a function which deals with
> the particular message, which , in my opinion is pretty neat. Comments?

An easy technique is to just do something like:

#define CMD_FOO 0x01
#define CMD_BAR 0x02

typedef int (*cmd_handler_fn)(struct server *svr, struct client *cli, cmd_t req, cmd_t resp);

struct {
char *name;
cmd_handler_fn handler_fn;
} cmd_tbl[] = {
{ "", NULL },
{ "CMD_FOO", foo_handler },
{ "CMD_BAR", bar_handler }
};

int
loop(struct server *svr)
{
cmd_t req, resp;
cli = select(...);
req = decode_msg(cli->fd);
cmd_tbl[req->cmd].handler_fn(svr, cli, req, resp);
}

Although I am sort of winging it at the end here it's probably a sound
design. If you need high performance you may need to think about how to
handler requests concurrently though.

Mike

[Martin Jost]

"Shashank Khanvilkar" <shashank@evl.uic.edu> schrieb im Newsbeitrag news:bgbflm$88q$1@newsx.cc.uic.edu...

> Hi,
> Will really appreciate for any comments on the below:
>
> Are there any Open Source tools that can convert a protocol design, as a
> Finite State Machine,
> directly to C code. (I can, of course, google for it, but I would like to
> get comments from people who
> have used such tools and how effective they are).

I know of libero [url]http://www.imatix.com/html/libero/[/url] , which can generate code for many different languages.
And are also Patterns in the GoF (Gang of Four) Book ("Design Patterns") which might fit.

HTH

Martin