TITLE> The York Linda Team - New Primitives </TITLE>

<H1> The Linda Team - Primitives </H1>

<hr>

We believe that the <TT>inp</TT> and <TT>rdp</TT> primitives as they stand are
semantically problematic (see <A HREF="pubs.html"> <I> Global Synchronisation
in Linda </I> </A> for more information). Therefore, we initially suggested the
replacement of them both with a new primitive <TT>collect</TT>. However recent
work has shown that <TT>collect</TT> replaces only <TT>inp</TT> and another
primitive <TT>copy-collect</TT> is required to replace <TT>rdp</TT>.
<P>
Both of these primitives require multiple tuple spaces. <P>

<hr>
<H3> THE <TT>collect</TT> PRIMITIVE </H3>

<TT>collect</TT> <I>moves</I> all tuples that match a given template from one
tuple space to another tuple space. It also returns the number of tuples
transfered. </P>

For more information see: <P>

<A HREF="pubs.html"> <I> Global Synchronisation in Linda </I> </A>
<hr>
<H3> THE <TT>copy-collect</TT> PRIMITIVE</H3>

The <TT>copy-collect</TT> primitive is similar to <TT>collect</TT> except that
instead of <I>moving</I> the tuples from one tuple space to another it
<I>copies</I> them. <P>

The <TT>copy-collect</TT> primitive solves a particular problem - which we have
come to call the <I>multiple rd problem</I>. This particualr problem invloves
many different processes trying to read non-destrutively a subset of the tuples
in a tuple space. Currently, it appears impossible to create an efficient
algorithm to do this. The <TT>copy-collect</TT> primitive appears to solce this
problem. There are a number of papers currently under development which
describe all this.....<P>

<P>
For more information see: </P>

<A HREF="pubs.html"> <I> Paper being currently written! </I> </A>

<hr>



<A HREF="linda.html"> Return to Home Page </A>