Two binary protocol questions (draft, stats)
aaron at serendipity.cx
Thu Apr 3 19:41:25 UTC 2008
On Apr 3, 2008, at 12:31 PM, dormando wrote:
> Brian Aker wrote:
>> On Apr 3, 2008, at 12:19 PM, Aaron Stone wrote:
>>> In any event, I think the stats key = value plain text response is
>>> completely reasonable.
>> If you look through the code you will find that quite a bit of
>> stats is undocumented for the protocol.
> That's partially by design; they weren't meant to be standard at
> time of writing.
> If we can deal with a key = value setup, we could return arbitrary
> values from arbitrary stats commands and still fit (enough) within
> the requirements of the current commands.
Oh, there are two issues: what command/response format to use, and how
to standardize the values?
Since stats is going to be used by automated monitoring systems, I bet
we do have to define and freeze the semantics of the stats values
(everyone now rolls their eyes at me, I just got this). So,
syntactically, key = value is fine, that's not the problem.
Semantically, I have no idea what the values are. Documentation time!
More information about the memcached