I would be willing to run such a plugin in production. Might want to have it
record incoming and outgoing separately. It might also be helpful for us to
see how many headers we are leaking...
-Bryan
On Jan 16, 2014, at 8:07 AM, Leif Hedstrom <zw...@apache.org> wrote:
> On Jan 16, 2014, at 6:34 AM, Igor Galić <i.ga...@brainsware.org> wrote:
>
>>
>>
>> ----- Original Message -----
>>> Updated Branches: cnf = TS_CONFIG_HTTP_CACHE_HTTP;
>>> break;
>>>
>>> + case 29:
>>
>> wow. Such expressiveness. I mean.. what's case 29?
>
>
> strlen() . Use the source, luke ;)
>
>>
>>> +++ b/proxy/hdrs/HdrToken.cc
>>> @@ -107,6 +107,7 @@ static const char *_hdrtoken_strs[] = {
>>> "Sender", // NNTP
>>
>> NNTP? Really? I thought we only support HTTP…
>
>
> Legacy, we could probably free this up. There’s a Jira on updating these WKS.
> Ideally, we’d collect metrics from real production systems, and make the WKS
> stuff based on some top % of request and response headers. Also, for the
> presence stuff, we should go through the source, since the presence
> optimizations are not public, only internal usage of presence benefits from
> the optimizations.
>
> Btw, I was thinking of a simple plugin, that listens on requests and
> responses, and produces simple metrics on header usage (not values). Would
> people consider running such a thing on 1 or a couple of boxes for some time,
> in production?
>
> Good stuff.
>
> — Leif
