Gervas Douglas (gmail) wrote:
>> If you have sequence numbers involved in your data exchanges, or you have
>> timeliness mechanisms that indicate missing data, and you can make an out of
>> band request to have the data resent, then you have transactional semantics
>> without software transactions needed. The software, end to end, can just
>> make
>> best effort to complete exchanges, and then you can rely on the two ends to
>> agree on the exchange.
>>
>> For those familiar with network protocols, TCP provides such end to end
>> guarentees without requiring every piece of the network, in between, to
>> guarentee data transfer. This is dramatic simplification of the hardware
>> systems involved.
>
> This is an interesting analogy. A connection-orientated protocol (or
> "oriented"
> if you use "orient" as a verb:) in comms is typically used at the level below
> which the protocol layers are judged to be reasonably reliable. So IP which
> occupies the Networking Layer
> (according to the OSI Reference Model) immediately below TCP's Transport
> layer,
> is connectionless. Where networks are less reliable error checking is done
> at
> the Networking or even Data Link Layers further down the stack.
>
> In contrast UDP, which also sits at the Transport Layer is connectionless on
> the
> assumption that the application takes responsibility for the error checking.
> Greg, does this analogy still extend to your scenario?
Right, years ago, reliable and unreliable transport were recognized as useful
features.
Gregg Wonderly
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/service-orientated-architecture/
<*> To unsubscribe from this group, send an email to:
[EMAIL PROTECTED]
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
