RE: Frames, Transactions, and Elements

Hi Per,

>Hi Shabtay,
>
>> [Shabtay] I'd like to see this EOT meaning end of transaction.
>
>The problem with using the term transaction the way you suggest is
>that it is too restrictive.  The EOM flag is optional.  You do not
>need to use it.  Some transactors may deal with fixed-size
>transactions and have a one-to-one correspondence between element
>and transaction.  Others may have variable length transactions, but
>may want to use a different mechanism than EOM for transaction
>delineation.  Some applications of the pipes may not even consider
>what they transport over these pipes transactions at all.
>
>So I think we need avoid using the term transaction.  It is an
>application-defined and dependent term.
>
>Per
[Shabtay] Whether we call it EOM or EOT is a nit. What is critical is
that it could be used to denote implicit synchronization points. EOM/EOT
could be used for accomplishing synchronization on transaction
boundaries even through it doesn't have to be (I agree with you on
that). I see no harm in using the term transaction as a composition of
one or more elements.