Hi Chuck,
Thanks for the feedback. I've made my code work with multipart
messages, and it is quite fast. It is quite flexible this way, so I'll
keep it :)


On Thu, Jun 2, 2011 at 3:14 PM, Chuck Remes <[email protected]> wrote:
> There is no reason to avoid using multi-part messages. I think they are a 
> great way to logically separate different components of a complex message so 
> I use them in that way pretty regularly.
>
> 0mq guarantees that all parts will be made available atomically. This means 
> that if you send a message with N parts, all N will be delivered or *none* of 
> them will be delivered.
>
> I have not seen any measurable performance issue with them. You could 
> probably write a simple benchmark to show what kind of overhead they incur. 
> If you need to serialize each message part separately (e.g. protocol buffers, 
> json, whatever) then the primary overhead is likely to be the invocation of 
> the serializer.
>
> cr
>
> On Jun 2, 2011, at 9:06 AM, Seref Arikan wrote:
>
>> Hi Martin,
>> Indeed this would be a solution, and Jashua has kindly provided a link
>> for a way to do this. I'm using protocol buffers in a complicated
>> wrapper scenario, and in general I'd like to get multipart messaging
>> up and running, since it feels as if it will offer more flexibility.
>> Given my current setup, multipart messages is method that offers
>> shorter time to functionality required.
>>
>> Any reason I should avoid multipart messages? Performance? Stability?
>>
>> On Thu, Jun 2, 2011 at 11:53 AM, Martin Sustrik <[email protected]> wrote:
>>> On 06/02/2011 11:37 AM, Seref Arikan wrote:
>>>>
>>>> I'm already using protocol buffers. The thing is, I can be passing
>>>> around any protocol buffer serialization; so I need an envelope to
>>>> tell the receiving party about the type of the message.
>>>> I'm using multipart messages to pass two protocol buffer
>>>> serializations; first one that contains data regarding the whole
>>>> message, the second one, actual payload of the message.
>>>> I'm happy to hear other approaches to handle this.
>>>
>>> Why not simply serialise both envelope and data using protocol buffers?
>>>
>>> Martin
>>>
>>>
>> _______________________________________________
>> zeromq-dev mailing list
>> [email protected]
>> http://lists.zeromq.org/mailman/listinfo/zeromq-dev
>
> _______________________________________________
> zeromq-dev mailing list
> [email protected]
> http://lists.zeromq.org/mailman/listinfo/zeromq-dev
>
_______________________________________________
zeromq-dev mailing list
[email protected]
http://lists.zeromq.org/mailman/listinfo/zeromq-dev

Reply via email to