[Asterisk-Dev] Dialplan syntax changes.. Option: work on a radically different design..

[Tzafrir Cohen]

On Wed, May 18, 2005 at 10:40:29AM -0400, Steve Kann wrote:
> CONSOLE=Console/dsp                             ; Console interface for demo
> ;CONSOLE=Zap/1
> ;CONSOLE=Phone/phone0
> IAXINFO=guest                                   ; IAXtel username/password
> ;IAXINFO=myuser:mypass
> TRUNK=Zap/g2                                    ; Trunk interface
> 
> 
> macro LocalDial (dialed) {
>    Dial($dialed);
>    switch($?) {
>    }
>    return something;
> }
> 
> context mycontext {
>    extension 2125551212 () {
>          Dial(SIP/Grandstream1,DIAL_RINGBACK|DIAL_ARGUMENT)

c-like, but with no clear distinction betwen strings and variables. Are
you sure this is still close enough to C syntax? A very common mistake
with editing isc-dhcpd's config file (uses c-like syntax as well, the
only example for such a program I could think of) is forgetting to quote
strings. Proper brace nesting is also a trouble-maker.

>          switch ($?) {
>             case NOANSWER {
>                2125551333 ();
>             }
>             case CONGESTION {
>                Play(busy);
>             }
>             default {}
>         }          
>    }
>    extension _212555XXXX () {
>          Dial(Local/$EXTEN);
>    }
> }
> ============
> 
> I think that we could get together and define a grammar for this, parse 
> it with yacc/lex, etc. 
> 
> There was lots of talk about the "pluggable PBX" concept earlier, on dev 
> chats, etc.  I think if you're going to break everyone's dialplans 
> anyway, you might as well really break them, and do it right.
> 
> It seems like what's happening here is that we're making the language 
> look like spreadsheet expressions..

That will never happen with C, as C has just about zero syntactic
support for decent data structures.

> 
> I'm surely oversimplifying the problem here, in at least two ways:
> 
> 1) The dialplan is not only looked at as calls pass through it, but it 
> is also consulted to determine the existence of extensions and pattern 
> matching.

Then why are we looking at procedural languages for inspiration?

> 
> 2) The dialplan was originally designed to be modifiable at run-time, 
> where you could, e.g. change things in the CLI, and then have asterisk 
> spit out the new dialplan.  But you can't do that now anyway.


Linux's iptables has somewhat similar requirements. It is a low-level API 
where rules are chained in chains. The API supports:

* creating/deleting chain/all-the-chains
* adding a rule to the end of the chain
* adding a rule to the beginning of the chain
* deleting a specific rule in a chain (identified by name or by number)
* replacing a specific rule in a chain
* "jumping" to a chain is a kind of a procedure call: if the packet has
  survived the chain it will continue from where it was

iptables had to remain low-level as it is implemented in the kernel and
the policy should be implemented in user-space.

-- 
Tzafrir Cohen     icq#16849755  +972-50-7952406
tzafrir.cohen at xorcom.com  http://www.xorcom.com