[Asterisk-Dev] IAX2 RFC and retries.

[Johnathan Corgan]

Steve Underwood wrote:

>> This is the sort of thing a formal protocol spec should clear up. 
>> (Bravo to the developer writing RFC--you'll improve a lot of peoples' 
>> productivity!)
> 
> 
> Was this said in jest, or was it a serious comment? If the latter, I 
> guess you haven't worked with too many specs. :-)

:-)

It was a serious comment though--spec writing is tedious, mostly 
thankless, and requires good writing skills *in addition* to the domain 
knowledge of the technical area involved.  So kudos to anyone who makes 
this effort.  They have my appreciation.

Besides, I'd rather read an RFC than decipher chan_iax2.c.  In 
particular, the differences between the MUST, SHOULD, and MAY items 
don't show up in the code.

Of course, the RFC writer is probably looking at chan_iax2.c and 
codifying what it does :-)

What will be interesting is when we have several interoperable but 
distinct open implementations of IAX.  I'm sure some bugs will appear 
that were hidden before because the chan_iax2 implementation is usually 
working with another copy of itself.  I haven't yet looked at the 
libiax2 source code tree yet, so maybe some of this has already been seen.

>> The timestamp does seem to be redundant as a way of correlating the 
>> ACK with the original full frame.
> 
> Why? The sequence number is entry you are expected to use.

This is what I was saying--comparing timestamps is redundant (and 
therefore unneeded), as one already has the source and destination call 
numbers and the sequence number in the ACK.

But both you and the other poster took my meaning to be the opposite, so 
I evidently didn't express myself well.

-Johnathan