[gambit-list] 0mq with Gambit

Feng Hou houfen at gmail.com
Thu Apr 7 19:46:59 EDT 2011


Sorry about previous top post, and this double post.

> Hi Feng,
>>
>> I'm glad to hear about your experiments and ideas.  Could elaborate on the
>> last comment that I've quoted above, re "not too keen on sending messages to
>> remote threads."   I think I must be misunderstanding something, because the
>> whole point of ZMQ is to send messages to remote threads.  I'm also not
>> clear what you mean by the terms subject-oriented message vs. too tightly
>> coupled. Feel free to elaborate on these, if you can.
>>
>> Thank you.
>>
>> Best regards,
>>
>> Jason
>>
>>
> Hi Jason,
>
> I meant to send messages to a specific thread _identity_. A receiver thread
> has to expose its identity (a pid or tid) on network in order to provide a
> service. Sender thread has to get hold of specific receiver thread's
> identity to ask for a service - tightly coupled (not much different from
> distributed object).This is useful model within local process, where shared
> state allow them to find each other easily. More important, message delivery
> is guaranteed, and peer errors can be handled easily. However, these
> properties become much difficult in distributed network. An architecture
> works better in distributed environment would be any threads in a cluster of
> gambit process nodes that can offer data or service on a subject just bind
> himself to a well known url (e.g. tcp://weather.com/va/reston/today, or
> pgm://stocktick.com/apple), can be one, two, or ten threads, they can come
> and go anytime as long as someone else take over their duty. Any threads
> consume services or data just send request message to well known urls, or
> simply bind their mailbox to them. Yes, ultimately 0mq delivers messages to
> gambit thread, but it decouples sender and receiver's identity.
>
> Hope this make some sense.
>
> Cheers,
> - Feng
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.iro.umontreal.ca/pipermail/gambit-list/attachments/20110407/d4ec6576/attachment.htm>


More information about the Gambit-list mailing list