[asterisk-dev] [PATCH] Silently drop non-conforming cirpack keep-alive packets

[Juan Carlos Castro y Castro]

Andrew Kohlsmith wrote:

>This was discussed on #asterisk-dev earlier this week.  Something like 
>"stupidvendorinterop=0x00000501", with each bit defining one stupid or odd 
>workaround needed for some vendor.
>  
>
Sounds nice, but somehow I doubt a 32- ot even 64-bit field won't 
eventually become not enough. :-/ Perhaps use descriptive strings like:

stupidvendorinterop=freetelecom-alienping
stupidvendorinterop=vonage-bye-weirdness
stupidvendorinterop=uzbekistan-broken-isdn

And have them resolve internally to how many bit fields as necessary.

(Disclaimer: the above should be not construed as an accusation of 
standards flaunting towards Vonage, Inc. or the country of Uzbekistan)

As a totally OT aside, I feel the same way about MSIE compatibility. 
Squid takes efforts to accomodate it; Firefox doesn't.

As a non-OT aside, I bumped into a IAX product that has broken IAX 
authentication when it has to be plaintext: it sends the plaintext 
password as if it's the MD5 password. The workaround I had to put into 
chan_iax2.c (in addition to the trivial change I suggested here, which 
is not an ugly hack) would be a good candidate for that entity too -- 
not useful if the rest of the world follows standards, but doesn't break 
anything or otherwise do harm.

I fully acknowledge this "accomodating bad behavior" thing is a larger 
philosophical discussion with no easy answer. I should know that; I'm a 
parent.

Juan