On Mon, 2005-06-13 at 15:06 -0700, Caitlin Bestler wrote:
> What's important is that enables dapl objects to be handled polymorphically.
> Internally there are many times when a utility routine only needs to know
> that something is a dapl object, not which one it is.
> 
> The alternative is to have some other mechanism to ensure that the
> shared fields are defined compatibly in all sub-types. This is doable,
> but declaring a common header is generally the simpler and more
> self-documenting method of implementing a parent type with
> multiple sub-types in plain C.

We have polymorphism already: dapl inherits from dat.  Are you saying
this is so we can have some sorta multiple inheritance?  I don't think
this is really used.  In fact, in dapl, the list is going away, the lock
(and flags) should be in a per struct anyways.  So, all we have left is
a pointer to ia?  Just add that to each struct.  Not a big deal.

> The external requirements for polymorphic dapl handles are fewer:
> get/set consumer handle (which ismostly be for user mode
> programs anyway)

done in dat, now.

>  and some methods apply to either a PSP or
> RSP.

which is a common structure with just a flag to tell the difference.

-tduffy

Attachment: signature.asc
Description: This is a digitally signed message part

_______________________________________________
openib-general mailing list
[email protected]
http://openib.org/mailman/listinfo/openib-general

To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general

Reply via email to