[Spread-users] Dynamic size scatter

Yair Amir yairamir at cs.jhu.edu
Wed Nov 10 06:36:07 EST 2010

Hi Heimo,

I think there is a misunderstanding how scatter / gather works.

The structure of the scatter is not really in the message. After the bytes
are "gathered" a message is sent as a contiguous entity to the other side.
At the receiver, if a scatter is used, that contiguous entity is split
according to whatever definition you have at the receiver, which can be
unrelated to how it was gathered at the sender. There is no record of the
sender scatter structure in the message that goes on the network.

This is just how scatter / gather works. It is an operating system service
and Spread just provides the same service, and uses the OS service where


	:) Yair,

On 11/10/10 4:05 AM, Heimo Zeilinger wrote:
> Hello,
> I was struggling with the same problem. I sent a scatter filled with n elements and expected to receive an identically structured scatter at the sink.  Unfortunately my expectation were not fulfilled and I decided for a workaround that equals more a dirty hack than a satisfying solution. Maybe I misinterpreted the scatter element structure but I expected that the scatter element size should be used at the receiver for decoding the element. 
> Maybe someone of you has an idea to solve this issue.  
> Kind regards     
> -----Ursprüngliche Nachricht-----
> Von: spread-users-bounces at lists.spread.org [mailto:spread-users-bounces at lists.spread.org] Im Auftrag von Baldr Helstad
> Gesendet: Dienstag, 09. November 2010 15:25
> An: spread-users at lists.spread.org
> Betreff: [Spread-users] Dynamic size scatter
> Hello everyone
> I've been trying to send and receive a scatter with scat_elements of variable length. At the receiving end I reserved the maximum of the possibly needed memory. 
> The problem is that SP_receive fills up the buf* with values until its full, ignoring the len values of the scat_elements.
> The question now is if it is possible to receive messages of variable length at all? 
> Sincerely, Stefan

More information about the Spread-users mailing list