[saga-rg] SAGA use case document: comments needed! (this week)

[Andrei Hutanu]

Let me rephrase this a bit. Reliability is good and useful as an option 
(as is in order delivery) but shouldn't be a requirement.
What this API should do is to remove the burden from the user to deal 
with arbitrary sizes of data  such as : 1 byte (current API) or 
1500/9000 bytes (UDP/RIPP) when using the network.
Regardless of wether this finds it's way into the SAGA API or not I 
would like to propose a phone conference
with all the interested parties to decide what this API should actually 
look like.

Andrei

Andrei Hutanu wrote:

> Hi John,
>
> I don't have reliability or in order-delivery as a requirement. The 
> idea is that there is a sender, sending large (variable in size)
> chunks of binary data.
> On the receiving end either a chunk comes complete (even if it has 
> some byte errors inside) or it doesn't come at all.
> Here is the tentative API I proposed (slightly modified from the 
> original use case here). Does this make sense?
>
> Start sending data asynchronously (cb gets called at the end of the 
> operation)
>
> AsyncResult beginSend (Block block,
>                       AsyncResultCallback cb,                       
> void* userData);
>    // releases resources, should be called in the callback, if called 
> before callback blocks
> endSend (AsyncResult handle);  
> Same as with send :
>      AsyncResult beginReceive (AsyncResultCallback cb, 
>                          void* userData);
>          Block endReceive (AsyncResult handle);
>    typedef struct Block {
>     const void* const buffer;
>     size_t            size;
> } Block;
>
>
>
>
> John Shalf wrote:
>
>>
>> On Nov 2, 2005, at 10:29 AM, Andrei Hutanu wrote:
>>
>>> Hi all,
>>> Here is my feedback regarding the Viz-LSU use case. In some cases
>>> these requirements might be outside the scope of SAGA, I would like  
>>> to know if that is the case though ..
>>>
>>> *Block-based data transmission is not covered by the current API
>>
>>
>>
>> By "Block-based data transmission" to you mean a kind of "Reliable  
>> Messages" API as opposed to the existing Stream API?
>>
>> If so, would you want to support more complex semantics such as the  
>> WS standard?
>> ftp://www6.software.ibm.com/software/developer/library/ws- 
>> reliablemessaging200502.pdf
>>
>> Or stick with simpler semantics like those offered by RUDP or RIPP?
>>
>>
>>
>