POST instead of GET for associate mode?

[Paul Crowley]

Brad Fitzpatrick wrote:
>>My vote would be to ignore MIME multipart, since generally you have to
>>jump through extra hoops to make it work on "web frameworks".
> 
> Could somebody note it in the spec?

Done.

I've made a few other changes to the spec (fixing a wrong name, noting 
the gaps).  The most substantive is that I've had second thoughts about 
what the reply to a "dumb consumer" should be: I want to use key:value 
pairs for this reply, so it's extensible.

The reason for *not* doing that is that there's only one value to 
return, and anything we can do to make dumb consumer's lives easier is 
good.  The reason *to* do it is my suggestion earlier: if a server gets 
a request that they can't directly honour - in particular, if there's a 
problem with the assoc_handle - then they can make a dumb mode response, 
which every consumer can handle.  In this instance, it would be good to 
use an extensible format for replies to check_authentication requests, 
so that they're useful to less dumb consumers too.

I don't think this is too burdensome to parse.  If you only want one 
field, parsing becomes something like

for line in file:
      if line.startswith("lifetime:"):
          return line.rstrip()[9:]

Sound OK?
-- 
   __
\/ o\ Paul Crowley, paul at ciphergoth.org
/\__/ http://www.ciphergoth.org/