On Sep 20, 2013, at 2:27 PM, Theo Schlossnagle <je...@omniti.com> wrote:
> experimental is experimental. No restrictions. Innovation comes more
> cheaply when breaking the rules and not complying :-)
Of these experimental plugins, I've starred the ones that I know to be deployed
in production:
* authproxy
balancer
buffer_upload
channel_stats
custom_redirect
* esi
geoip_acl
* healthchecks
hipes
lua
memcached_remap
metalink
mysql_remap
* rfc5861
spdy
stale_while_revalidate
* tcp_info
I'm ok with "unstable" being unstable, but let's not be blase about it :)
> On Fri, Sep 20, 2013 at 4:29 PM, Leif Hedstrom <zw...@apache.org> wrote:
>
>> On Sep 20, 2013, at 11:24 AM, Phil Sorber <sor...@apache.org> wrote:
>>
>>> Asked about this in IRC, but wanted to bring it to the list. For
>>> experimental plugins, do we need to abide by the backward compatibility
>>> rule in a stable release? Same question probably applies to
>>> ts/experimental.h.
>>>
>>> My opinion at first was no. It's experimental, duh! But the more I think
>>> about it, I think it does apply. We want people to feel comfortable using
>>
>> I agree with this, if it's experimental, it ought to be ok to change
>> behavior. Moving plugins to stable seems good too, people need to champion
>> this for the plugins they wrote and/or care about (such as using it for
>> real traffic).
>>
>> -- Leif
>
>
>
>
> --
>
> Theo Schlossnagle
>
> http://omniti.com/is/theo-schlossnagle
