directory-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Emmanuel Lécharny <elecha...@gmail.com>
Subject Re: Replication configuration
Date Wed, 14 Nov 2012 06:11:26 GMT
Le 11/14/12 7:01 AM, Kiran Ayyagari a écrit :
> On Wed, Nov 14, 2012 at 11:24 AM, Emmanuel Lécharny <elecharny@gmail.com> wrote:
>> Le 11/14/12 6:16 AM, Kiran Ayyagari a écrit :
>>> On Tue, Nov 13, 2012 at 3:00 PM, Emmanuel Lécharny <elecharny@gmail.com>
wrote:
>>>> Hi guys,
>>>>
>>>> I have some questions about the replication configuration...
>>>>
>>>> AFAICT, we may configure the server to be a consumer, a producer, both
>>>> of them, or none of them.
>>>>
>>>> A standalone server will not be a producer or a consumer
>>>> A master server will be a producer
>>>> A slave server will be a consummer
>>>> A master/slave server will be a producer and a consumer
>>>>
>>>> In any case, a server will always be able to be a producer as soon as it
>>>> allows another server to connect on it as a consumer, and get
>>>> replicated, so for a server to be a producer is just a matter of
>>>> receiving a consumer request.
>>>>
>>>> Am I right ?
>>>>
>>> yes
>>>> I'm asking because I added a flag that allow a server to become a
>>>> producer on demand as soon as the ads-replicationEnabled flag is set to
>>>> true (this flag is used to start the replication handler when the server
>>>> is started). Does it makes sense to keep this flag ? Or should we always
>>>> start the replication handler ?
>>>>
>>> prior to this flag we have always looked for the presence of
>>> ads-replReqHandler attribute
>>> if it exists we instantiate this class and the server acts as a provider/master
>>>> I'm inclined to think that teh flag is a good thing to have. wdyt ?
>>>>
>>> If the above attribute's absence can be treated as a 'false' value so
>>> IMHO this flag is spurious in a way
>> The reason I added it is that without it, we will either require that
>> the user add the replication handler FQCN (not very simple to find it),
>> or to start the handler automatically using the defaumt value. I don't
>> find both solution very convenient.
>>
> agree, I just fancied the idea of switching replication implementations
> based on custom protocol implementations
>> I would rather question the fact that the replicaton handler is loaded
>> using its FQCN : why don't we just have a boolean flag, and nothing else
>> ? (if the flag is true, we laod the default handler)
>>
> absolutely, we can go ahead just with the flag except if we want to still allow
> injecting a new provider impl
Then that means we need the flag and the replication handler AT. This is
currently what we have.


-- 
Regards,
Cordialement,
Emmanuel Lécharny
www.iktek.com 


Mime
View raw message